11  Scholarly Literature

(Modified on 2026-08-11)
2,542 words

When starting a research project, the first step is doing background research on the topic you are interested in. We need to learn what work has been done on our topic, to make sure your research question, theories, hypotheses, and protocols make sense.

As you find and read literature, you should start to write your paper, especially the introduction and background sections. Write your ideas (theories, hypotheses, research plans) and note what ideas you need to improve or cite, and go looking for those citations. If you come across a paper using a cool methodological idea (an algorithm or an experimental design) that you want to use, write that part of your methodology section saying “We use the xxx methodology (Smith, 2012).” Sometimes you’ll find another source which gives you a better idea, and you’ll rewrite your ideas a bit, using that citation. This is an iterative process; keep it up and you’ll eventually flesh out a solid research plan for your project.

As you find and read background literature, you’ll find that your research idea changes as you take into account past work you didn’t know about before. You may even feel discouraged and decide to drop the project entirely. This is fine of course (), but don’t feel discouraged if you find out that your idea has “already been done.” This is not a bad thing! It means you are free to answer the next question, or zero-in and refine the existing work. Actually, replicating previous work can be extremely useful and valuable in itself!

Types of sources

For serious research papers, all or nearly all citations should come from scholarly, peer-reviewed “primary sources.” Primary sources include

  • Journal articles
    • The principal publications in the sciences and social sciences.
    • Example:
      • Arthur, C. (2021).
        Vicentino versus Palestrina: A computational investigation of voice leading across changing vocal densities.
        Journal of New Music Research, 50(1), 74–101. https://doi.org/10.1080/09298215.2021.1877729
  • Proceedings articles
    • The principal publication in engineering fields.
    • Example:
      • Arthur, C., & Condit-Schultz, N. (2023).
        The Coordinated Corpus of Popular Musics (CoCoPops): A Meta-Dataset of Melodic and Harmonic Transcriptions.
        Proceedings of the International Society for Music Information Retrieval
  • Monograph books by academic scholars
    • The principal publications in traditional arts and humanities scholarship.
      • Humanities professors are expected to publish whole books, not “just” articles.
    • Example:
      • David Huron’s Sweet Anticipation (2006), which is widely cited in music cognition.
  • Chapters in collections (book)
    • These books are essentially collections of original articles on a similar theme, published together. Their peer review is less standardized, but they can still be high impact articles.
    • Make sure your reference for these mentions the book editor(s) and the chapter author(s).
  • Other journal formats (lower impact):
    • Letters to the editor.
    • Commentaries or responses (to other papers).
    • Data reports.
    • Book reviews.
Most research articles cite few or no secondary sources.

Sometimes, some pieces of knowledge are so general and fundamental that you can’t find a primary source that even mentions them. In these cases, you can cite a “secondary source.” Secondary sources include encyclopedia entries, textbooks, and “handbooks” which are written by active researchers but focus on reviewing and summarizing existing work. Good examples would be the Psychology of Music (2013)), a collection edited by Diana Deutsch but with chapters contributed by many prominent researchers in music psychology. Secondary sources are appropriate sources for “basic facts” or “big picture” ideas about your topic, but only if you can’t find a primary source that mentions them. For example, if you want a citation for the claim that “the 4/4 time signature indicates the most common type of duple meter,” you might go to Oxford Music Online (an encyclopedia). In practice, you’ll find that most research articles cite few or no secondary sources.

Another type of scholarly source is work by active academic researchers which isn’t properly published, or is published but not peer reviewed. For example, self-published preprints are papers that have not been peer-reviewed (yet)—in fast-advancing engineering fields, preprints have become frequently cited. Another example would be conference abstracts, proposals, summaries, or even proceedings papers that are published online but which are not actually (fully) peer reviewed. Sometimes you might find an abstract for a talk given at a conference 15 years ago which is totally relevant to your work, but there is no paper to read! You can cite this, but since you don’t know the content of the talk, the citation must be fairly vague, mainly serving to show that you are aware the work exists. Of course, if you actually attended the talk, you can cite it more confidently; still, it’d be better if you can find a published (peer-reviewed) paper to cite in addition.

Note that, confidently citing something like a talk or a non-peer-reviewed paper as if it is a strong primary source might be interpreted as deceptive. Either use non-peer-reviewed work to support minor points or make sure to “frame” them in a skeptical way. You want your main important claims and ideas supported by peer-reviewed work.

Non-scholarly sources

Generally, all citations should be “scholarly”—published by known academic journals, societies, or publishers, and usually authored by academics (though some industry researchers publish as well). Still, citing non-scholarly work may be appropriate or necessary in some cases. For example, it will always be difficult to find academic citations about recent trends in popular music—if you are researching a trend that started last year, you might need to cite a newspaper article or blog post. However, you should always try to find something scholarly to cite. Ideas from non-scholarly works shouldn’t really be cited as an authority, but more like anthropological observations. For example, I wouldn’t want to cite a blog post like this:

Music performed on instruments tuned to \(432_{Hz}\), instead of the standard \(440_{Hz}\), is more attuned to human brain waves and evokes more relaxed feelings (SomeBlobPost, 2022).

However, I might say some thing like:

The theory that \(432_{Hz}\) tuning is inherently more pleasant to the human brain has been widely circulated on social media and blog posts (see for example, SomeBlogPost, 2022).

I’m not accepting the scholarly source as authority, but using it as an example to support my claim.

To be fair, many newspaper articles and even blog posts are well written by intelligent people, perhaps even incorporating data analyses, etc. You might want to cite data/results/ideas from an intelligent non-scholarly article, but don’t try to “pass it off” as (peer-reviewed) scholarship.

Types of articles

The vast bulk of your citations in a research paper should be research articles (from journals or proceedings). Three types of research articles can be distinguished: the research report, the meta-analysis study, and the review article.

  • A research report communicates the results of some original research. This is the standard form of scholarly article; Perhaps 95% of all articles in the sciences and social sciences are research articles.
  • A review article is a more theoretical work that summarizes many studies related to a particular topic and attempts to present a coherent interpretation about what all of the research implies.
    • For example, Juslin and Laukka’s (2004) article reviewing research on music and emotion. Sometimes a review article does the reverse: it may draw attention to the incoherent and contradictory state of research pertaining to some topic, and may offer advice for future research that might help to resolve the mess.
    • Most review articles are well-written and highly informative. They provide a good place to start when you begin working on a new project. Unfortunately, review articles are rarely labelled “I’m a review article.” Some journals, like Psychological Review specialize in publishing review articles. Because review articles are popular with readers, commercial publishers often aim to produce “Handbooks,” where each chapter reviews research on a specific topic written by a different author.
  • A rarer type of article is a meta-analysis study—a “study of studies.” Meta-analyses are typically done when a large number of studies have been carried out related some problem.
    • For example, many studies have been carried out related to whether television violence promotes violent behavior in viewers. Over a hundred studies have been carried out on this topic. Some of the studies seem to show a link, whereas other studies seem to show no link. In a meta-analysis, the researchers identify all of the pertinent studies. They then evaluate the quality of each study, including the quality of the samples used, the number of participants, the quality of the stimuli, the extensiveness of the controls, and other factors. Poor studies are simply discarded if they fail to achieve the minimum quality criteria established by the researchers. Then the researchers combine together all of the good studies, and do a statistical analysis on the aggregate data. The aim is to see if all of the studies ultimately tell a coherent story. In critically reading any meta-analysis, the main concern is how the author(s) set the criteria for including or excluding a study in the final data analysis. Does the exclusionary criteria seem sensible?
    • In general, meta-analyses are rare in music research. However, a fine example is found in the work of Friedrich Platz and Reinhard Kopiez (2012), who examined the influence of visual information in assessing the quality of a musical performance.

Finding literature

If you know the article(s) you want and just need to access them, its often easiest to find them either on the web (google scholar) or through the GT library. The library has access to various databases. If you can’t access something you need, ask your research advisor for help—we might be able to access it somewhere, or we can request access through the library. Searching the web/library for keywords sometimes works, if you are lucky, but often doesn’t work well.

The harder task is looking for articles that you don’t know about. Your best bet is to look in specific places that are likely to have what you want:

  • Primary publications (e.g., journals, proceedings) that are important in your area of interest:
  • Secondary publications related to your topic:
  • Ternary publications related to your topic:
    • Pop-science books (good ones include citations).
  • Directly from the researchers themselves!:
    • The personal websites of researchers or websites of labs/research groups.
    • Networking:
      • Go to conferences and talk to people.
      • Reach out to researchers via email (its a good idea!).

Spend time getting to know the various sources that are most relevant to you. That includes journals, books, conferences, etc. but also people: who are the researchers working on your topic? Whenever you read a paper you like, note who the authors are, where they work, and where they published. Talk to your colleagues and advisor about what sources they use, and researchers they watch.

Begin with articles you already have and look up the (relevant) papers they cite.

The ultimate technique to find unknown citations, is to begin with articles you already have and look up the (relevant) papers they cite. Then, for each of those citations, look at everything they cite, and for each of those, what they cite, etc. ad nauseum. This can take a long time, and can be exhausting, but there is no better approach.

When searching for articles, you’ll have to decide which articles to actually access so you can read them. You should decide on which sources to target based on:

  1. How they are cited in something else you’ve read.
  • This is the most important one. If other work you care about cites a source, you should probably read that source.
  1. Which journal/source it is published in.
  • Work in high-impact sources is usually worth reading.
  1. The title and keywords.
  2. The abstract.

Note that the title and keywords (3.) don’t provide you very much/good information but they can be quickly read and scanned.

Reading articles

There are a lot of articles out there. Reading them all can take a huge amount of time. Using efficient reading strategies is essential.

Skimming

When searching for citations, you will come across many articles that may or may not be relevant to you, to various degrees. Some papers you will obviously need to read through in full. However, a lot of the time you need to skim a paper quickly, so you can decide whether you want to read for more details or find citations in the paper that might interest you more. For the most part, you should skim a lot more papers than you actually (fully) read.

The key to “skimming” is that you don’t read the whole paper—you only read as far as seems relevant/interesting and then you move on. In each paragraph, read the first few sentences until you get the gist of the paragraph; if the paragraph is highly relevant to you, read the rest, otherwise jump to the next paragraph. Once you’ve skimmed the first few paragraphs of a section, you might jump to the next section. As you do this, you may realize that you’ve missed something, and need to jump back. That’s fine, this process doesn’t need to be linear.

An effective skim through a research paper might look like this:

  1. Read the abstract.
  2. Read the Introduction as far as you need to get the relevant gist.
  • The first few paragraphs of an introduction are often “big picture”; Jump through them quickly to get to the end of the intro (the narrow part of the funnel where they might state the hypothesis or summarize the results. If the paper still seems interesting and valuable, you can go back and skim/read through the introduction in more depth.
  1. Next, jump forward to the start of the Conclusion section, searching for more statements of their main results/findings.
  2. If you are interested in particular details that you still can’t follow, search for those details in the body of the paper.
  3. If the paper is clearly highly relevant to you, put it in your queue for a thorough reading.

At any step in skimming, you might decide the paper is not useful to you and stop. At this point you might simply forget about the paper and move on but you should at least glance through the references section first to see if there are any other sources you might want to find and skim.

The basic idea of skipping is to jump around between parts of a paper where important information is usually located (end of the intro, beginning of the conclusion). However, when skimming a paper, we often know what we are looking for. For example, here are important things you might want to know about:

  • Is this an experiment, an exploratory study, or a correlational study?
  • What is the hypothesis being tested?
  • What data is used and how is it sampled?
  • What are the stimuli?
  • What are the independent variables?
  • What are the dependent variables?
  • What are the algorithms or models used?
  • What variables are controlled?
  • How is the result evaluated?
  • Is the result strong, weak, or negative?
  • What are the conclusions?

In most cases you are skimming a paper because there is something particular in that paper that you are interested in. If so, you should zero in on scanning for that information—this might be part of step (4.) above. For instance, maybe there is a methodological detail, like the experimental design, that you might want to use in your own work—so skip to the methodology and scan for that information.

Full Reading

Note, skimming and scanning is part of the process of finding literature to cite. For any papers that you actually end up citing, you should read them thoroughly to make sure your citation makes sense. It’s possible that, if you are just citing a paper for a very specific (minor) point, you might not need to read the whole paper For example, if you are saying something like “transformer models have been used extensively in object recognition tasks (Smith 2021, James 2023, Marcus and Pierce 2020)” you probably don’t need to know every detail of these three papers. You just need to be sure that they all use transformer models for object recognition tasks. If, for example, James (2023) started out planning to use a transformer model but decided it didn’t work and decided to switch to an adversarial model, that would be a bad citation—so read enough to be confident you have interpreted the citation correctly.

Take-Home Messages

When we finish reading a research article, our natural tendency is to make a mental note of the conclusion. We want to distill what we’ve read into some sort of “take-home message.” For example, after reading an article by Smith and Jones, we may try to remember that they showed (say) that lower pitch is heard as sounding sadder. Good researchers try to avoid such simple distillations. The sign of a careful researcher is that they remember a few important details about the study. Instead of simply recalling “low pitch is heard as sadder,” they will recall something like: “in a two-alternative forced-choice experiment, American non-musician listeners report that the melody with two or three pitches lowered by a semitone sounds sadder.”

Research is difficult enough without being burdened by ideas that are wrong. It’s not what you don’t know that is most problematic: it’s what you think is right (but is wrong) that causes the greatest havoc. You never know when it will be appropriate to reinterpret a past result. Instead of showing “low pitch sounds sadder,” you may later realize that a better interpretation is “lower-than-normal pitch sounds sadder,” or “higher-than-normal pitch sounds happier (less sad).” Research results may depend on the specific question. For example, a researcher may get different results if they ask participants to judge “unpleasantness” rather than “dissonance.” So it may be important to recall that experiment X measured “unpleasantness” whereas experiment Y measured “dissonance.” Conversely, many judgments are highly correlated. Researchers may find they get the same results whether they ask listeners to judge “sadness” or “sleepiness.” There are always other ways of interpreting the results of an experiment.

For important experiments, good researchers will recall the dependent and independent variables used and the way that the terms of the hypothesis were operationalized. They’ll also remember whether the effect size was large or small. Finally, they’ll pay attention to whether other researchers have replicated the result.

Note that there might also be useful take-home messages for you that aren’t necessarily the main message of the paper. For example, you might note take-homes like

  • “this particular experimental design doesn’t work for what I want to do because of X, Y, Z.”
  • “the loss function they use would be perfect for my experiment.”
  • “I should access this dataset (that they are sharing) for my own project.”

Make sure to take notes and annotate your bibliographies!

Further Information

If you have a question about some published article which you can’t figure out from the paper itself, don’t be afraid to write to the author.

Professional editors typically want authors to be brief and to-the-point. This is enforced in proceedings papers by the short maximum page length (4–6 pages). Even in journals, editors sometimes ask authors to shorten or summarize an otherwise verbose description. Consequently, articles may omit various details about the research. If you have a question about some published article which you can’t figure out from the paper itself, don’t be afraid to write to the author. Most researchers are happy to answer questions about their work.