Colleagues, PhD and master’s students alike may pay some attention to the following. I had just come across of publisher Springer’s announcement of former colleague Alexandrina L. Dumitrescu’s latest book “Understanding Periodontal Research”, see here. The book has yet not been published but it is assumed that it will be available in the end of November. A link by Springer leads to a chapter on “Multilevel Modelling in Periodontal Research.” As the “Abstract” tells,
“Periodontal data are usually plentiful observations made in one oral cavity. In order to describe the periodontal situation, sites (gingival units) around teeth within patients or subjects are considered by using several variables. Observations may be even repeated in a longitudinal way. This is a typical hierarchical situation with lower and upper levels.”
As some readers may know, I had dedicated considerable time and effort in the past ten years to multilevel modeling of site-specific data in periodontal research which has resulted in so far 13 scientific papers and quite a number of conference abstracts on the topic. On my Institute’s webpage, I have had announced years ago,
“I am pretty much convinced that site-specific clinical observations are most valuable and of utmost importance in the understanding of periodontal diseases. Unfortunately, in recent years, these observations have frequently been summarized at the subject level. While doing so, much of the collected information is lost. Since site-specific analysis of periodontal data may yield an amazing deal of new information regarding the pathogenesis of periodontal disease, I would invite anybody who owns site-specific data, which has been conventionally analyzed before with the subject as statistical unit, to share his/her data in order to do a correct, i.e., multilevel analysis.”
Well, Springer’s “Abstract” of Dr. Dumitrescu’s chapter (without quotation) sounded quite familiar to me. It took not too long that I found the article of which these sentences had been borrowed. In 2008, I had written a brief tutorial on “Dealing with hierarchical data in periodontal research” which had been published in Springer’s dental journal Clinical Oral Investigations . I next googled these four sentences and immediately found most pages of the respective chapter on Multilevel Modelling in Periodontal Research at GoogleBooks. My four sentences in question were not an abstract. They were rather necessary part of an introduction which leads to the formulation of the aim of the paper, namely how to deal with hierarchical data in periodontal research (I will turn to that below). Well, Dr. Dumitrescu’s used them out of context, verbatim, as her Introduction (with no further ado) to the chapter (p. 297), though with a correct reference to my original article, Müller (2009). What is missing here are the quotation marks .
Copy and Paste
Now I became really interested. GoogleBooks not only provides most of the multilevel modeling chapter but many more pages of the whole, yet not published, book. My very first impression (well, hypothesis) was that Dr. Dumitrescu had copied and pasted all sentences, even paragraphs, conclusions, pictures and tables from other sources, without bothering of timely interpretation or any further intellectual input, just quoting her sources but not using quotation marks. It is tempting to test this hypothesis but, since GoogleBooks does not allow printing out the pages which are displayed by permission of the copyright owner Springer, this is going to be cumbersome (I had meanwhile contacted Springer for the original manuscript).
The next three paragraphs in Dr. Dumitrescu’s chapter which stretch over two printed pages (with heading 11.1 Site-Based, Tooth-Based, and Subject-Based Periodontal Evaluation, and subheading 11.1.1 Clinical Perceptions) refer to a recent review article about “Site-based versus subject-based periodontal diagnosis” by G. Rutger Persson . Dr. Dumitrescu starts with a somewhat awkward sentence, which turns out to be verbatim one of Persson’s conclusions in his article (p. 155),
“Presently, the pattern of disease progression defines the primary differentiation between the chronic and aggressive periodontitis. The current periodontal diagnostic system disregards age as a determinant in a differential diagnosis of periodontitis.”
Without concluding the paragraph here, Dr. Dumitrescu continues with a couple of sentences which can be found verbatim in Persson’s introduction (p. 145).
“The differential diagnosis between chronic and aggressive periodontitis requires examinations at more than one single time point. It is clear that the subject is the unit of observations for both conditions. […] In contrast, most periodontal intervention studies, and clinical periodontal routines, have established standard examination procedures that rely predominantly on site-based. […] Most clinical periodontal trials using the periodontal site as the unit of observation combine dichotomous (i.e. bleeding on probing, plaque scores, and tooth mobility) and numerical observations (i.e. probing depth measures, and clinical attachment loss). Such data are then often used independently (e.g. site bleeding on probing, with pocket probing depth greater than a defined threshold value), and are not merged to get information on the subject level. In the clinic, dentists make diagnostic, prognostic and therapeutic decisions using tooth or site as the unit of observation, automatically absorbing dichotomous and numerical information in the diagnostic process (Persson 2005).”
Under the then following sub-subheadings Site-Specific Diagnostic Methods and Subject-Based Periodontal Diagnosis, Dr. Dumitrescu just lists verbatim (and refers to without quotation marks) Dr. Persson’s respective summaries in his own article, including identical bullets. There is no need to repeat them here. Under the subheading 11.1.2 Statistical Issues follows verbatim just the abstract (except its last sentence) of a 1990 paper on the assessment of the relationship between site-specific variables by Hujoel et al. .
A search for the next two sentences in the papers to which they are referred  were largely unavailing. Dr. Dumitrescu writes,
“Sites are clustered around a tooth, and teeth are clustered in individuals. It is therefore inappropriate to analyse the site-or subject-level observations using single-level, univariate (sic!) statistical methods because the correlation among sites and/or teeth within an individual invalidates these statistical methods (Albandar and Goldstein 1992; Hoberman 1992; Wan et al. 2009).”
Similar sentences can only be found in the rather late paper by Wan et al. where the usage of multilevel modeling in a treatment study is justified in its respective Introduction (p. 230) and its Material and Methods section (p. 231). While the paper by Albandar and Goldstein (1992) has been an early application of multilevel modeling to periodontal data, the comment by Hoberman (1992) on the contrary attempts to deny the periodontal site as proper statistical unit. Reference to the paper by Tu et al. (2004) in her next sentence, “Consequently, statistical analysis with the assumption that the site observations are independent would generate potentially misleading results” cannot be found in the respective paper and has to be regarded an erroneous quote. Tu’s paper (with coauthors)  deals with mathematical coupling without relation to multilevel modeling. I will come back to that below.
So, one has to conclude that both introduction and first section in Dr. Dumitrescu’s chapter on multilevel modeling were entirely composed of copied and pasted sentences from the given sources, verbatim and without any modification. Quotation marks are missing, though. “Original input” which seems to occur in the final three sentences of the first section is spurious and garnished with erroneous quotations.
Copy from Methodological Papers
And so it continues. Needless to tell that the two introducing paragraphs of the second section, i.e. 11.2 Principles of Multilevel Modelling (MLM) and the next paragraph with the subheading 11.2.2 The Advantages of Using MLM in Periodontal Research are copied verbatim without quotation marks from Gilthorpe et al. (2000); and 11.2.3 The Disadvantage of Not Using MLM in Periodontal Research and 11.2.4 Example of a Multilevel Model from Axtelius et al. (1999) , including all formulas and unusual notation of variances.
I have tried, to no avail though, to identify a source for the following section 11.2.5 Types of Effects which is a naive and nonsensical attempt to explain what is meant by fixed and random effects. Dr. Dumitrescu erroneously refers to another, in essence narrative, paper by Gilthorpe et al. (2000)  who give an overview on the potential of multilevel modeling in the analysis of dental data and which is the first and introductory article in a series of related papers. In section 11.2.6 The Application of MLM to Dental Research Data, Dr. Dumitrescu refers to respective papers in this series of the early 2000s and newer articles, explicitly stressing that the respective authors had introduced multilevel modeling to dental research but at the same time ignoring previous studies dating back well into the 1980s, for instance Sterne et al. (1988) . Eventually Dr. Dumitrescu returns to my own paper’s introduction (Müller 2009) ) when concluding (on p. 300), “A nontechnical description of the basic principles of multilevel modelling and, as an educational example, the application of multilevel models (the random intercept and random coefficients model) to the data on the thickness of facial gingival (sic!) were recently presented,” with a correct reference (emphasis as such) .
Section 11.3 deals with The MLwiN Software, which is just copied and pasted verbatim and without quotation marks from a 2002 article by one of the gurus of multilevel modeling, Harvey Goldstein, and members of the Center of Multilevel Modelling which is right now located at Bristol University, who were essentially involved in the development of this particular, user-friendly specialist software for multilevel modeling .
Copy and Paste Results
The following major sections of the chapter are dedicated to, yes, results. In 11.4 Application of Multilevel Modelling to Cross-Sectional Periodontal Data, a paragraph (p. 224) in the aforementioned paper by Gilthorpe et al. (2000) is modified by omitting the context in which it was formulated (repeated measures in longitudinal studies). Then, paragraphs of a large number of Results sections of original articles with respective interpretations and conclusions by the authors are just copied and pasted, verbatim without quotation marks. This even includes certain original tables and figures of which the publishers’ reprint permission had been obtained (but what about reprints of entire paragraphs?).
A focus of Dr. Dumitrescu’s interest seems to be mathematical coupling and related regression towards the mean which does not have a relation to multilevel modeling. A paper by Tu et al.  again serves to provide the reader with a short introduction, just by copying and pasting verbatim, even including all references cited by the authors. Results, including several figures, of a certain paper on mathematical coupling by Tu et al. , is particularly featured. It appears especially absurd when a paragraph contains a reference to an “Appendix” and a dead link to the journal’s web page where the article had been published.
Not surprisingly, Dr. Dumitrescu’s chapter on multilevel modeling ends with what it has started. For her conclusion Dr. Dumitrescu picked two sentences verbatim (without quotation marks) from the final paragraph of my tutorial (Müller 2009) :
“[…] the main advantage of multilevel modeling in periodontal research is the possibility of unbiased dealing with the true, hierarchical, structure of the data. […] New study designs should take into consideration the tremendous power of these techniques.” 
A New Concept?
Remains the question why Dr. Dumitescu dedicated a lengthy chapter about a certain advanced statistical method such as multilevel modeling in her book on “Understanding Periodontal Research”? Well, given the complete lack of any critical appraisal of the different models presented in the numerous studies she mentions; the lack of any discussion of authors’ interpretations of findings; and a general lack of own ideas and concepts, Dr. Dumitrescu’s approach is in fact anti-scientific. It may well be that this chapter in her recent book (in a series of five, all published by Springer) is only the tip of the tip of an iceberg. Writing “Understanding Periodontal Research” may have been, well, an ego trip. But who actually wants to follow her own long way of comprehension?
In one of the first of her books Dr. Dumitrescu writes (as can be read on its back cover ),
“Periodontitis is a complex, multifactorial disease and its susceptibility is genetically determined. The present book systematically reviews the evidence of the association between the genetic variants and periodontitis progression and/or treatment outcomes. Genetic syndromes known to be associated with periodontal disease, the candidate gene polymorphisms investigated in relation to periodontitis, the heritability of chronic and aggressive periodontitis, as well as common guidelines for association studies are described. This growing understanding of the role of genetic variation in inflammation and periodontal chronic disease presents opportunities to identify healthy persons who are at increased risk of disease and to potentially modify the trajectory of disease to prolong healthy aging. The book represents a new concept in periodontology with its pronounced focus on understanding through knowledge rather than presenting the presently valid answers. Connections between genetics and periodontology are systematically reviewed and covered in detail.” (Emphasis added.)
But this is not a new concept. Understanding through knowledge, that is what, in the old days, students had to try, often with rather questionable results since accumulation of knowledge does not automatically facilitate a critical mind. Presenting the presently valid answers is no new concept either. What makes science vital is challenging presently valid answers.
Now, What Can We Learn From This?
A quick look at right now available parts reveals that her so far described method of arrogation of other scientists’ intellectual property may apply to other chapters of Dr. Dumitrescu’s book as well. I have to concede that I would not have paid attention if my current area of interest including my own work had not been a prominent issue in the book. Since an estimated 90 per cent of the text, which I have analyzed by now, has been copied verbatim from published papers it comes well-written, and often ideas are reasonably developed in a logic way. Change in style in a piece of scientific work is frequently an indicator for possible plagiarism. However, it seems that, due to editors’ demands, medical and in particular natural scientists have, in contrast to scientists of humanities and arts, adopted over the years an almost indistinguishable and essentially prosaic style with, in a certain field, also frequent use of universal jargon. So, there was in Dr. Dumitrescu’s case apparently no need for making any adjustments to a certain personal style (if it existed).
Systematic investigation of violations of minimum scientific standards in doctoral dissertations including shameless plagiarism has recently led respective Universities to withdraw doctoral titles and in some cases to more or less forceful resignation of a number of politicians in Germany. Prominent individuals include, among others, Germany’s Defense Minister Karl-Theodor zu Guttenberg, members of the European Parliament Silvana Koch-Mehrin and Jorge Chatzimarkakis and political scientist Margarita Mathiopoulis. The doctoral thesis by Germany’s present Minister for Education and Science (sic!) Annette Schavan is right now under scrutiny for plagiarism as well.
What plagiarism (from Latin plagiarius, kidnapper, seducer, plunderer, one who kidnaps the child or slave of another) actually is seems to be clear as standard dictionaries mention, for instance, the wrongful appropriation, close imitation, purloining and publication of another author’s language, thoughts, ideas, or expressions, and the representation of them as one’s own original work . Does providing the source as reference abrogate plagiarism? In most cases, yes; in particular when unmistakably indicating the quote with quotation marks. Certainly not if virtually all sentences, paragraphs, sections had just been copied without any discernible own intellectual input as in the present case.
Plagiarism in academics or journalism is considered dishonest and a breach of professional ethics with sometimes grave occupational consequences such as expulsion from an academic institution or other severe career damage. In particular, manual proof of plagiarism is frequently impractical, but even computer-assisted detection employing currently available software may be not as easy as usually believed.
As the entire book will be published soon, I shall continue to report on this case as appropriate.
 The context is a brief explanation to an unfamiliar reader of why data are to be analyzed in a different way when they have (as often is the case) a hierarchical structure. So, the first sentence of my Introduction, “How to statistically analyze collected data very much depends on its structure” is an inevitable part of developing an argument in the following respective four sentences which eventually lead to the aim of the paper. Taking just the respective four sentences verbatim out of context is not sufficient as an introduction to a whole chapter, which deals with the very topic; and even less is it an “abstract” as her publisher wants to suggest on its website.
 Hujoel PP, Loesche WJ, DeRouen TA. Assessment of relationships between site-specific variables. J Periodontol 1990;61:368-372. As not being a member of the American Academy of Periodontology, Dr. Dumitrescu had probably no access to the full paper (no access through the University library either) which is a valuable source as regards assessing the respective association in the old days, for instance by applying procedures which had been described by Mantel and Haenszel. A highly useful textbook is in that regard the chapter on “Combining evidence from four-fold tables” in Fleiss JL. Statistical methods for rates and proportions. 2nd ed. J. Wiley & Sons; New York 1981, pp. 160-187. Reading once again after so many years Dr. Hujoel’s paper of 1990 (with Walter Loesche and Timothy DeRouen), which even refers to early attempts of applying advanced statistical methods such as multilevel modeling and the Generalized Estimating Equations to dental datasets, reminds me of having actually been at the cutting edge of site-specific analysis of periodontal data for more than two decades.
 Albandar JM, Goldstein H. Multi-level statistical models in studies of periodontal diseases. J Periodontol 1992;63:690-695. Hoberman D. Periodontal sites or patients as experimental unit. J Periodontal Res 1992(Spec Issue);27:417-419. Wan CP, Leung WK, Wong MC et al. Effects of smoking on healing response to non-surgical periodontal therapy: a multilevel modelling analysis. J Clin Periodontol 2009;6:229-239. The glitch in Dr. Dumitrescu’s choice of “univariate” rather than uni-(or rather single-)level may reveal her complete unfamiliarity with the subject. The topic is not about univariate/multivariate analyses here but how to deal with the hierarchical structure in the data.
 Tu YK, Maddick IH, Griffiths GS, Gilthorpe MS. Mathematical coupling can undermine the statistical assessment of clinical research: illustration from the treatment of guided tissue regeneration. J Dent 2004;32:133-142.
 Gilthorpe MS, Griffiths GS, Maddick IH, Zamzuri AT. The application of multilevel models to periodontal research data. Community Dent Health 2000;17:227-23; Axtelius B, Söderfeldt B, Attström R. A multilevel analysis of factors affecting pocket probing depth in patients responding differently to periodontal treatment. J Clin Periodontol 1999;26:67-76. When copying from Axtelius et al. (1999), few editorial changes had been made. A reference to Albandar and Goldstein (1992) was removed while those to Bryk and Raudenbusch (1992), Claffey and Egelberg (1981), Cooley et al. (1981) and Woodhouse et al. (1995) were just kept as they were provided in Axtelius’ (with coauthors) paper. Axtelius’ “[…] 2 principal objections to uni-level analysis when a hierarchical structure is present” were collated. PPD was changed to PD.
“The aim of this communication is to give a rather non-technical description of the basic principles of multilevel modeling and, as an educational example, the application of multilevel models to data on the thickness of facial gingiva.”
As Dr. Dumitrescu rightfully notes, several increasingly complex extensions, beginning with a variance components, or null, model to a random intercept and random coefficient model, were developed in the tutorial which are unfortunately not dealt with at all in her own chapter on multilevel modeling in periodontal research.
 The final paragraph in  actually reads (p. 277):
“As a first conclusion, the main advantage of multilevel modeling in periodontal research is the possibility of unbiased dealing with the true, hierarchical, structure of the data. While correct estimates of fixed effects are obtained, the random part of the model (variances, covariances) may provide new and deeper insights into phenomena and mechanisms at the level of interest, the periodontal site. New study designs should take into consideration the tremendous power of these techniques. Multilevel modeling should not only be applied in an attempt to simply obtain the correct (unbiased) fixed estimates, for example, in a situation where in a few cases multiple observations in certain patients were recorded. In general, random effects should not be considered as nuisance as is the case in more commonly used marginal models [reference to Diggle et al. 2002]. They are not ‘errors’, but valuable information regarding data structure. Random effects should therefore properly be reported, since they may reveal new ideas about subject and tooth variation, correlations between covariates, and sometimes even problems with the chosen model. In particular, collaboration of professional statisticians and clinical researchers is urgently needed to significantly improve clinical research.” (Emphasis added for the purpose of this review.)
 About plagiarism’s very limited legal implications, see, for instance Stuart P. Green’s article in Hasting’s Law Journal here. Interesting in particular his position on plagiarism and the question of mens rea (intent).
“Perhaps some of the confusion about the moral status of plagiarism can be attributed to a deeper confusion about the mental element, if any, necessary for its commission. Some ethical codes prohibit only “intentional” or “knowing” plagiarism. Others prohibit plagiarism that is either “intentional” or “unintentional”— that is, they treat plagiarism as a kind of “strict liability” offense. Finally, a large number of codes (surely, a majority) prohibit unattributed copying without specifying what, if any, form of mens rea is required. Moreover, as we shall see, most newspapers and magazines do not even have a written code of conduct. Thus, when some people think about plagiarism, they are assuming that it requires an intentional or knowing act. Others are thinking about conduct that is inadvertent, though perhaps recklessly or negligently so.
Should plagiarism require an intent to deceive or some other mental element, or should it be viewed as a strict liability offense? I would argue that, just as morality informs law, so too should law inform morality. If theft requires intent, and plagiarism derives much of its meaning from theft law, it seems to follow that plagiarism should also require intent. At the same time, I would modify this requirement to say that the element of intent can be satisfied by “deliberate indifference” to the obligation to attribute. That is, if the reason a person was unaware that he was copying or failing to attribute is that he was deliberately indifferent to the requirements of attribution, he should be viewed as having committed plagiarism.” (Footnotes omitted, emphasis added.)
Massive copy and paste including even all references in the original paragraphs of quoted articles point not only to deliberate indifference to the requirements of attribution, i.e. numerous quotation marks throughout, in the present case. Most probably unable to either comprehend or digest, and discuss what she read the author trusted the experts and stringent peer review by established scientific journals. So far, I have not received a response by Springer’s editor as to whether Dr. Dumitrescu’s work had been peer reviewed.
22 November 2012 @ 8:07 am.
Last modified November 23, 2012.