There is No Scientific News. The failure to determine what we already know due to our poor literature searching practice
Faced with an anarchistic database of four million psychological articles, psychological researchers’ searching techniques seem woefully inadequate to find the state of the art. It is our academical duty to do far better. Literature searching is a difficult and essential skill of ‘slow science’.
Most scientific news is not news. Not in the sense of factual news items such as ‘pro-Trump mob storms U.S. capitol’ or ‘women in Swiss military no longer forced to wear men’s underwear’. Scientific findings, except for straightforward technological inventions and some archaeological findings, need multiple replication and reviews which takes a lot of time and turn news into ‘olds’. In the public’s eyes it will have become outdated, expired, stale, eventhough it may have, at that point, become truly newsworthy. But even then: how do we know that it is really new? How clear is the state of the art to which the finding adds something new? To suggest our cutting-edge results, we researchers typically use phrases such as “… it is not yet known that…” or “… there is no evidence yet that…” often after no more than a sketchy review of the literature. We hardly ever report our exact literature search. Even if we do, for example in review articles, we rarely provide exact keyword phrases that others can check by repeating it in the same database. Still, should a review not be replicable? Even that of any empirical article? Are we not ethically obligated to show that our findings are really new and contributing to science? That we are not reinventing the wheel, fooling ourselves and the rest of the world? Science and the general public should not let us researchers get away with inadequate overviews based on superficial searches with loosely formulated lists of keywords.
But is it so bad? Are we really so poor at rendering the state of the art in our own fields of research? I have reason to believe so. In 30 years of doing my ‘peer review duty’ I have seldomly read a really good, systematic search in the literature in the introduction of empirical articles, nor even – albeit somewhat better - in reviews and meta-analyses. At the many conferences I visited hardly any lecture started with a truly comprehensive rendering of the ‘what we knew’ before presenting the findings purportedly being ‘what we did not yet know’. Admittedly, my own reviews used to be pretty sketchy too. When my students and I started taking it more seriously we discovered that good literature searching is astonishingly difficult. Initially we thought that this was due to our topic of study, stress, which is notorious for its hopeless definitory issues, with even hundreds of keywords still failing to completely cover the concept. But over the course of several years, we discovered that this appears to be true for any concept in psychology more or less. Now, you might think that in your particular field the ‘obvious keywords’ suffice for an easy and quick literature search and even monthly updates. The reality however seems to be that ‘obvious keywords’ are rarely enough to cover concepts in psychology and related sciences. For almost every concept a multitude of terms are used, and this continues to grow. While scientists in general delight in labelling concepts and phenomena, psychologists have an anarchistic terminological history that has led to a virtual Babylonic confusion of tongues. Walter Mischel suggested that theories in psychology are like toothbrushes: everyone wants their own, and no one wants to use others’.
Let's take a few examples. Emotion is often discussed using a multiplicity of terms such as affect, feeling, mood or motivational factors, or many specific emotion words are used. For good old unconscious there is an astonishing wealth of keywords, as we found out: inducing implicit, unaware, automatic and many more. Even for a relative newcomer in psychological research, mind wandering, terms are rapidly increasing, daydreaming obviously but also task-independent thoughts, task-unrelated thoughts, self generated thoughts, stimulus independent cognition, and so forth. Terminology related to willpower has proliferated too: long versus short-term reward, delay discounting or temporal discounting or more recently even ego depletion. Placebo is often referred to as non-specific effects, context effects, expectancy, and so on. Even famous cognitive dissonance find itself under other headings such as expense justification. The exact combination of words also matters a great deal. For early life stress I constructed and tested 147 different permutations of phrases like early life stress, childhood adversity, trauma in infancy, before finding out that circa 25 were sufficient for my particular question (its relationship with heart rate variability in adulthood).
My point here is that building such a keyword profile is an important academic skill in itself, and is woefully underestimated. For many keyword profiles it took my colleagues, students and myself weeks or even months to construct them. In a series of short video lectures, I explain why literature searching practices are really poor, and hamper establishing the state of the art for our scientific questions (see BOX)
There is a second, perhaps even bigger problem that severely hinders determining the state of the art in one’s field, namely the overwhelming and exponentially growing number of scientific articles published every day: about three peer-reviewed per minute (!). In psychology alone, about 1 peer-reviewed article every 3 minutes (there are now more than 4 million peer reviewed articles in psychology alone). In my own specialty, stress, one article is published every hour. In such an oceanic volume of articles, again only very precisely specified keyword profiles can help us to keep abreast in one’s own area.
So, dear colleagues, I’d like to challenge you now. Are you still convinced that you know exactly the state of the art in your own specialism? Do you really use all possible keywords and combinations (not to mention sufficient truncations, the correct OR’s and ANDs and brackets etc. - in short, your keyword profile) for your topic, whether it is conflict, creativity, happiness, memory, identity, group processes, love, self etc.? Is your search replicable by other researchers? Will I find exactly the same when I use your keyword profile, and thus at the same state of the art?
Let’s face it: demonstrating exactly how and why a finding is new is the ethical obligation of every scientist. An article without adequate state of art misrepresents truth and novelty. It also leads to chronic under citation, and a failure to give appropriate credit to other researchers. The average social science paper is cited a small fraction of one citation per year, and the majority are never cited. Poor citation practice is largely the result of poor literature searching and is causing academic self-harm, hampering the accumulation of knowledge and reinventing the wheel again and again. Emeritus professor of psychology Albert Kok (in Dutch): warned not so long ago that the disappearance of the historical continuity of the field gives the false impression that our PhD students almost always conduct ground-breaking research. And we are all responsible for this.
I hope that I was able to convey the importance of using advanced literature searching techniques with a comprehensive and replicable keyword profile, and that the latter is a difficult and truly academic skill. I have taught this skill for many years to undergraduate as well as graduate students, and the aforementioned video clips (see BOX) came forth from this. For many psychology students it will perhaps even be the most important practical scientific skill to learn since most are not going to do research themselves, while their later jobs will appeal to their skill of finding scientific knowledge. But that skill should also be further developed in senior scientists, and should become ubiquitous across science. I also propose that not only theoretical review articles, but also any article reporting empirical work should start with a much more exact review of what we already know and what we don't know yet, and that review should be replicable. Isn’t a review a study in itself and shouldn’t it therefore be controllable by others? Thus, it should contain a precise keyword profile, and any other author should get the same overview using this profile, or change it and publish it as such.
Will this slow down science, this thorough reviewing, even for every single empirical article? Yes, it will be a slow process. But it is needed to discover and rediscover what has been reported before in those many millions of articles, amongst which those that have never even been cited. As the late Dutch psychologist Piet Vroon, one of the last who tried to become a ‘homo universalis psychologicus’, remarked at the end of the last century: we should stop experimenting for the next 100 years and read what has been published in the previous 100 years.
So: slow science indeed. But science that provides true and lasting answers to questions. A true accumulation of knowledge. It will take longer. Too long perhaps to become ‘news’, but … still new!
If you wish to find out more about the problems of literature searching for state-of-the-art reviews and the solutions I propose, please view the videos and the PDF with additional detailed tips . I also deal with the wrong solutions, such as using books or – surprising for many – Google Scholar, looking only in the last 5 years, etc. And the solutions, including how to test multiple alternative keywords or keywords phrases in a very efficient and quick way. How to find keywords. The quirks but also advantages of Web of Science over other databases. And much more. Finally, you will learn how to become the true specialist on your topic! And isn’t that what society asks from us?