If anyone needs convincing that proliferation of supplemental materials to journal articles has become a real problem they only need to look at the following chart (Fig. 1) to be disabused of their innocence.

Fig. 1

Journal of Clinical Investigation. Percentage of published research articles with supplemental material per year . Chart courtesy of Ken Beauchamp, American Society for Clinical Investigation. Used by permission

Over the past decade, not only the number of articles with supplemental materials has been increasing steadily but the growth of supplemental materials size has been exponential (Fig. 2).

Fig. 2

Average size of a Journal of Neuroscience article and supplemental material. Source: Maunsell, J. (2010), Announcement regarding supplemental material, J Neurosci 30(32): p.10599. Used by permission

In his article [1], Todd Carpenter referred to supplemental materials as “Pandora’s box of issues.” Let us look at the contents of this box. In fact, supplemental materials comprise a diverse set of objects, including but not limited to:

• multimedia;

• gene sequences, protein structures, chemical compounds, crystallographic structures, 3-D images;

• computer programs (algorithms, code, libraries, and executables);

• text, tables, figures (constituting entire sections of the articles, such as Materials and methods, Extended methodology, Survey results, Bibliographies, Derivations, etc.); and

• datasets.

It needs to be emphasized that datasets, though important, are not the only type of supplemental materials.

Making supplemental materials available in electronic form has provided unquestionable benefits to the scholarly community: authors can share supporting evidence in the form of multimedia and datasets to corroborate their conclusions; they can present in-depth studies that are not available in print; reviewers can evaluate the research more thoroughly; and readers can replicate experiments and verify the results.

However, the veritable explosion of supplemental materials has posed a number of practical challenges, such as:

• Degree of importance. If an article has a number of supplemental materials, are they all of equal importance? As a busy reviewer or a reader already suffering from information overload, how can you determine which supplemental objects are critical for the article’s science and which are merely "nice to look at?"

• Discoverability. If you just skim through the article or deal only with its alternate representation (e.g., its bibliographic record, abstract/response page, or an entry in the table of contents), can you always determine whether the article contains supplemental material?

• Identification and linking. If you chance upon orphaned supplemental material online, how can you ascertain what its parent article is? And how do you cite supplemental material or provide a persistent link to it?

• Viability. Will it be possible to render (read, play, execute, etc.) supplemental materials in 20 years? 200 years? Given the rate of software and hardware obsolescence, it is likely that in order to ensure longevity, supplemental materials will have to undergo constant conversion.

• Preservation. If forward migration of supplemental materials is inevitable, how do we ensure that the supplemental object that the future user interacts with is equivalent to the original one? Should the original be preserved?

• Transmission and packaging. When an article with supplemental materials is transferred, e.g., from a publisher to an archive or from a library to a reader via an interlibrary loan, how do we ensure that nothing was lost or corrupted in transmission?

• Intellectual property rights. Who has rights over supplemental materials, and where are those recorded?

• Curatorial responsibility. Who has custody over supplemental materials: an author, a publisher, a library, a data center, an institutional repository, an archive, or/and any other actor?

• Business models. Last but not least, if someone is going to provide identification, description, preservation, and other types of processing of supplemental materials, what are the sustainable business models that would support that?

As we can see, the issues posed by handling supplemental materials affect practically all actors participating in scholarly communications:

• Authors

• Editors

• Peer reviewers

• Publishers and their own or contracted hosting platforms

• Data centers and institutional repositories

• Abstracting and indexing services

• Citation indexing and reference linking services

• Libraries

• Archives

• Historians of scholarship

The community has responded to the challenge in various ways. Researchers appear to be split on the issue: while some argue that more supplemental materials should be made available and are optimistic about the technology’s ability to solve some of the above problems, others argue that a scholarly journal "is not a data dump" and an article "is not an FTP site."

Different publishers too responded in different ways. In 2009, Cell imposed strict limits on the number and kind of supplemental materials that could be accepted [2]. In 2010, The Journal of Neuroscience banned supplemental materials altogether and announced that it would embed dynamic content in its articles’ PDF format [3]. In 2011, The Journal of Experimental Medicine introduced a policy limiting supplemental materials only to "essential supporting information [4]."

Among several efforts that preceded the formation of the NISO/NFAIS Working Group on Journal Article Supplemental Materials two especially are worth mentioning. In February 2009, NFAIS issued Best Practices for publishing journal articles [5].

One key recommendation on Supplemental Materials was that the journal make a clear connection between an article and the supplemental materials that accompany it. Once published, the supplemental materials should be considered part of the journal’s archival record and should not be changed without a clear statement of correction. Publishers, the document noted, should always supply a recommended citation as well as good, descriptive metadata for those materials. A&I services covering the journal article should include the presence of supplemental data in the article record, indicating file types and DOI [6].

In October 2009, Alexander (Sasha) Schwarzman […] conducted an informal survey of scientific journal publishers to learn how other publishers were dealing with the issue of "supplemental materials." Conducted mainly through the e-mail listservs of CrossRef TWG and eXtyles, Schwarzman's questions "touched a raw nerve" and generated more responses than he had been expecting. [...F]ull survey report [was] issued in November 2009. [...] Schwarzman's article was the impetus for a January 2010 Supplemental Materials Roundtable meeting on the subject co-sponsored by NISO and NFAIS and the subsequent Working Group on Supplemental Journal Materials that the two organizations launched [7].

In August 2010, the NISO/NFAIS Working Group on Journal Article Supplemental Materials began its activity.

The Working Group consists of a well-balanced mixture of publishers, librarians, archivists, A&I and citation linking services, and independent consultants. They come from the private sector and government, commercial and non-commercial, academic and non-academic institutions.

The intended result of the Working Group's activity is to produce Recommended Practices for supplemental journal article materials. The Business Working Group (BWG), co-chaired by Linda Beebe and Marie McVeigh, concentrates on the business policies and practices, such as:

• scope of and general principles for Recommended Practices;

• definitions: supplemental materials, article, data, metadata, multimedia;

• curation and life cycle: selection, peer review, editing, production, presentation, providing context, referencing, citing, managing/hosting, discoverability, and preservation;

• intellectual property rights management and accessibility; and

• roles and responsibilities of authors, editors, peer reviewers, publishers, libraries, A&I services, repositories.

The Technical Working Group (TWG), co-chaired by David Martinsen and Alexander ('Sasha') Schwarzman, focuses on the technical aspects of handling supplemental materials, such as:

• metadata, persistent identifiers, and granularity of markup needed to support practices recommended by the BWG;

• referencing and linking to and from supplemental materials, handling cited references within them;

• archiving, preservation, and forward migration of supplemental materials;

• packaging, exchange, and delivery of supplemental materials; and

• technical support for accessibility practices recommended by the BWG.

In addition to the BWG and TWG, there is a Stakeholders Group that provides feedback on draft documents and community vetting of a final document. The group list is open; anyone who would like to track the progress of this project and provide feedback can sign up by visiting: www.niso.org/lists/suppinfo.

Perhaps the most significant Working Group’s conceptual contribution in the supplemental materials discourse is a clear distinction it has introduced between Integral and Additional types of supplemental content.

Integral Content is not really supplemental; rather, it is essential for the full understanding of the work by the general scientist or reader in the journal's discipline but is placed outside the article for technical, business, or logistical reasons.^* Examples of Integral Content include descriptions of methods needed to evaluate a study, review, or technical report; detailed results required to comprehend outcomes; tables, figures, or multimedia with primary data required to verify or fully understand the work. In general, the publisher maintains responsibility for hosting and curating this content in the same way the article itself is treated. (For some specialized journals, content held in an external repository may be considered integral.)

Additional Content is truly supplemental: it provides a relevant and useful expansion of the article in the form of text, tables, figures, multimedia, or data. It may aid any reader to achieve deeper understanding of the work through added detail and context. Examples of Additional Content include expanded methods sections; extended bibliographies; additional supporting data or results; copies of instruments/surveys; and multimedia and interactive representations of additional, relevant, and useful information. Generally, the author has created this content, and the publisher hosts it or places it on the open web.

Related Content. The Working Group spent considerable time discussing other content the author wishes to make the reader aware of because it may add to the understanding of the work, or aid the replication or verification of the results. Examples include data, gene sequences, protein structures, digital recordings, 3-D images, and chemical compounds used, created, or deposited by authors and held in external repositories. Related Content generally resides in an official data center or institutional repository. The author may or may not have been the creator, and the publisher has no responsibility or authority over this content and does not host it. Journals treat references to this content differently. Many expect it to be listed as another cited reference, and others link to it outside the citation list. Because the publisher lacks any authority, the Working Group offers no recommended practices for that content.

The difference in Integral and Additional Content goes far beyond the distinction between those things that can and can not be printed. A non-essential table can be printed, but it is still Additional (i.e., truly supplemental). While an essential video can not be printed but still remains Integral—there is nothing supplemental about it except the adjective used to describe it. The fact that an essential object is not embedded into the article because of technical, business, or technological reasons should not obscure the fact that it is still essential. Such content can be referred to as “pseudo-supplemental”—it is treated as if it were supplemental while, in fact, it is anything but. Thus the term "supplemental" is, in fact, a homonym: it refers to both the Additional (truly supplemental) and the Integral ("pseudo-supplemental") content.^**

Table 1 provides a summary of BWG recommendations.

Table 1

Summary of the BWG Recommendations.

Developing a robust metadata schema is the foundation for addressing all aspects of the TWG charge. The TWG has developed the first draft of a schema that includes descriptive and physical metadata, relationships between supplemental materials and a core article, and relationships among supplemental objects. There are a number of conceptual and practical challenges that TWG still needs to address.

• Heterogeneity. A supplemental archive (ZIP, TAR, RAR) or a document (PDF, MS Word) may contain both Integral and Additional Content, which may present a problem, especially in identification and linking

• Relationships. Some objects are logically different but have a few metadata elements in common (e.g., a series of aerial photographs), while other objects constitute alternate representations of the same logical entity, e.g., multiple representation of a protein or of a chemical structure

• Recurrence. An archive or a document may contain nested objects and groups

• Hierarchies. An archive may contain a tree with many branches and sub-branches

• Granularity down. At what level should metadata be applied: the entire set of supplemental materials, object groups, or individual objects?

• Granularity up. Should the set of supplemental materials be linked to the core article as a whole or should individual supplemental objects or their groups be linked to a specific item within the core article, e.g., a figure or a table?

• Related Content. Should it be marked up, and if so, how extensively?

• Integral and Additional Content. What is the extent of the differences in marking up Integral and Additional Content?

Even though the Working Group has stated that business models are beyond the purview of the Recommended Practices it is developing, the TWG is very conscious of the fact that assigning the metadata and applying markup, coupled with implementing the associated quality control procedures, is going to result in real costs for publishers without discernible benefits at this stage. Thus the TWG is trying to find a balance between comprehensiveness and practicality of the metadata schema and to build flexibility into how each publisher may choose to implement it.

When the metadata schema for supplemental materials is developed and adopted, a question will inevitably arise of how to integrate it into the existing Tag Sets, such as JATS. This does not appear to be too difficult: JATS already contains a wrapper element supplementary-material that can encompass such typically supplemental objects as graphic, media, or table. Perhaps, a structural section needs to be added to the list of elements that can be found inside supplementary-material wrapper, so that the entire section of a paper, such as Extended Methodology could be tagged as supplemental. Then, a parameterized list of supplementary-material's attributes could be extended to incorporate the metadata developed by the Working Group.

What is currently missing from JATS, however, is a mechanism for indicating the degree of importance of supplemental materials. To distinguish between Integral and Additional Content, a number of possibilities exist: for example, an existing attribute @specific-use could be used for that purpose, or a new dedicated attribute @importance could be added to the supplementary-material element.

It is worth noting that JATS as it exists now can already accommodate articles in the journals that do not treat essential dynamic content as supplemental,^* e.g., if a version of record for the article contains an integrated video or 3-D image, JATS already allows a media object media or a chemical structure wrapper chem-struct-wrap to be part of a paragraph in the article's body.

In her Editorial, Cell's Editor-in-Chief Emily Marcus laid out her vision:

… over time the concept of supplemental material will gradually give way to a more modern concept of a hierarchical or layered presentation in which a reader can define which level of detail best fits their interests and needs.

If that comes to pass, the binary model of Integral versus Additional Content may, perhaps, become replaced by a more nuanced one, resembling tree rings, where each supplemental object or group is assigned a weight, not unlike the mechanisms employed in the relevance ranking algorithms.