1. Discovery depends on establishing that a problem is significant enough to be labeled an important achievement.
The cognitive psychology of science literature has little to say about this; most studies either give participants a task to solve, or use standard problems already accepted as significant by the discipline.
Future cognitive research should study situations where scientists have to make decisions about which research to pursue and also develop laboratory simulations that permit participants to choose among a set of scientific tasks. Robert Rosenwein and I have outlined one way to accomplish this sort of research program (Gorman & Rosenwein, 1995). (I will say more about this in Section 5.2).
2. Discovery depends on transforming that problem into a form that suggests a promising path to solution which includes locating and transforming the necessary data.
In terms of locating data, the literature on dual-space search is helpful, suggesting that a careful, coordinated search of both hypothesis and experiment spaces is most likely to uncover evidence that is relevant to the problem at hand. The literature on expert-novice differences gives us hints as to how experts transform textbook-style problems: they try to classify them based on the underlying principles required to solve them, rather than the surface features. The result is the kind of abstraction used by Galileo to solve pendulum problems.
More research needs to be done on what happens when experts move out of their familiar domains. Here work on analogies may be especially helpful. Local analogies appear especially useful in research teams, allowing mental models and techniques that are useful in one domain to be transferred to a closely-related domain. It may be that many more discoveries are made by this kind of local analogy than by remote analogies, though--as we saw in the section on analogies--there are prominent examples of analogies that seem remote playing an important role in discoveries.
Consider, for example, Darwin's famous insight after reading the sixth edition of Malthus's Essay on the Principle of Population.--"the polemical account of humanity outstripping its food supply, and the weak and improvident succumbing in the struggle for the available resources." (Desmond & Moore, 1991) Darwin already knew Malthus' theory, but it was the statistics in the sixth edition that convinced him that "A struggle for resources slowed growth, and a horrifying catalogue of death, disease, wars and famine checked the population. Darwin saw that an identical struggle took place throughout nature, and he realized that it could be turned into a truly creative force." (Desmond & Moore, 1991).
Darwin made an analogy between the human struggle exemplified by the poor laws and pauper riots, and the struggle between species, an analogy that certainly seems remote, if not distant. But as Darwin admitted, when he read Malthus, he was already "well prepared to appreciate the struggle for existence which everywhere goes on from long observation of the habits of animals and plants" (Gruber, 1981). Part of Malthus' argument was that the weak would succumb in the human struggle for increasingly scarce resources. Similarly, Darwin saw that favorable variations might triumph in nature. Darwin's experience and background knowledge allowed him to take an apparently remote analogy and transform it into a local one.
3. Discovery depends on a combination of flexibility and stubbornness, depending on the cognitive styles and career trajectories of the scientists involved and on how they represent the problem.
The literature on confirmation and disconfirmation suggests that, in general, a scientist ought to begin by trying to confirm any hint of a pattern that might lead to a promising hypotheses, seeking positive results. This positive test heuristic might end up falsifying the pattern, or discovering that there is a great deal of noise in the data. However, the scientist should not deliberately seek disconfirmation until she has found a pattern or hypothesis that merits this kind of rigorous scrutiny.
Of course, scientists do not operate in isolation. One strategy is simply to propose bold hypotheses for which one has amassed some confirmatory evidence, and let others attempt to falsify it. One particularly persuasive cognitive style involves arguing consistently for a novel hypothesis while at least appearing to consider potentially disconfirmatory results (Rosenwein, 1994). Prominent discoverers in psychology like Freud, Skinner and Simon embarked on extensive research programs, took at least some results critical of their perspectives into account, but never abandoned their 'hard core' ideas. It would be interesting to see whether such a cognitive style is equally effective in other sciences.
4. The act of writing is part of the discovery process.
There is a literature on cognition and writing which includes protocols of writers (Flower & Hayes, 1984). Similar protocols ought to be done on scientists as they write. Berkenkotter and Huckin (Berkenkotter & Huckin, 1995) describe how one biologist wrote and revised an article, and Myers (Myers, 1990) describes how two other biologists wrote and revised grant proposals. These works provide a valuable catalogue of the rhetorical moves and revision strategies followed by the biologists, but do not include the kind of fine-grained analysis represented by a protocol.
For example, Berkenkotter & Hucking followed the process by which their biologist published a research article in a refereed journal. In her own words, the biologist had to learn to tell a 'phony story' about how her research was conducted. In her early drafts, she constructed a narrative based on the internal logic of her own research program. In later drafts, she made it sound like her own research had emerged from issues in the scientific literature in her area. I followed a similar trajectory when I revised one of my experimental research articles (Gorman, 1992a). These accounts of scientists learning the conventions of their genre are very valuable, because, as Holmes documented, discoveries can emerge when a scientist transforms her research results for publication. As part of her quest for publication, the biologist studied by Berkenkotter and Huckin ended-up generating new data which confirmed her general approach--but it might also have generated a surprise that led to a discovery.
Protocols should also be done on research teams as they compose articles. One heuristics for collaborative writing involves delegation, with a senior scientist establishing a title, focus and organization and farming out specific sections to junior members. At another extreme, the collaboration among members of the team can be so intimate that by the time an article is published, it is impossible to say who was responsible for the overall plan of the piece. It would be interesting to know how Dunbar's teams wrote the articles that announced their results.
5) Successful discoverers often pursue a network of enterprises.
Here, aside from Gruber's seminal workd (Gruber, 1989) the cognitive science of science has little to contribute--the kinds of simulations, computational or experimental, that have been done have tended to focus on single problems and problem-solvers. But there is no reason why cognitive psychology could not be applied to following a scientists' network, as Gruber did, looking closely at decisions about what research projects to pursue and the links among the different research enterprises. Such a network can be a great source of the kind of analogies that lead to new discoveries, as we saw in Darwin's reading of Malthus. Darwin's network included detailed studies of barnacles, pigeon breeding, the way in which bees made honeycombs, geological stratification and species variation in the Galapagos. His network helped make the Origin of Species a juggernaut of persuasion.
What is needed in future studies, is the systematic application of some of the research findings discussed in this chapter. Gruber's provocative work on Darwin would profit from the kind of fine-grained protocol analysis that Tweney and Gooding have done on Michael Faraday. One could, for example, make inferences about how Darwin employed confirmatory and disconfirmatory heuristics as he coordinated a search in hypothesis and evidence spaces. Gruber's own framework was Piagetian, which means he focused on assimilation and accommodation, where assimilation involves fitting evidence into existing conceptual structures and accommodation involves altering conceptual structures. A confirmatory heuristic is a way to accomplish assimilation, and a disconfirmatory a way to encourage accommodation.
But to really study the interplay of these heuristics, one needs to follow their use at a finer-grained level than Gruber does. Instead, what he provides are provocative hints. For example, he notes that Darwin had developed a theory of the formation of coral reefs by 1835; this theory contained many of the important features he would use in his theory of evolution two years later, including the idea that population growth is a struggle against natural forces and that different species of coral adapt to different environments. But he did not at this earlier date see the parallels that seem 'obvious' in hindsight. Gruber 's account of the reasons why Darwin did not see the analogy is incomplete.
We need a more detailed study of the interplay between hypothesis and evidence in the coral case to see what kinds of heuristics and mental models Darwin could have transferred to the evolution case. This appears to be another situation where Darwin had to make an analogy that initially seemed remote into something that was obvious, given his background and experience. In this case, the analogy is from one species to all species; he clearly needs to see that the case of coral is not peculiar and can be linked to other cases like the Galapagos finches. One could imagine constructing a computational simulation, perhaps using CLARITY, to explore various possible relationships between hypotheses and evidence in this case. One could extend such a simulation to cover other aspects of Darwin's network of enterprises, and perhaps provide a model for generalization to other scientists' networks.
In defense of Gruber and others who study historical cases, records are not always available to do the equivalent of a protocol analysis and/or a detailed computational simulation. One solution to this problem is to study modern scientific research teams, like the ones followed by Dunbar. The difficulty here is the daunting amount of data that is produced. One solution would be to encourage those who study discovery in vivo and those who simulate it in vitro and on computers to collaborate. Funding agencies like the National Science Foundation and the Spencer Foundation could certainly encourage this.
The brief review of the cognitive psychology of science literature conducted in this chapter indicates that it makes an important contribution to several of the generalizations above, particularly numbers 2 and 3. But our review includes at least as many calls for future research as summaries of existing findings. Cognitive psychology of science suffers from the fact that it is a recognized area of interest for only a very small group of researchers. Hence, it is not really a research area, but a diverse group of 'invisible colleges' (Crane, 1972) pursuing their own research programs, often with to little connection to each others' work or the other science studies disciplines. If nothing else, it is hoped that this chapter will encourage more of a synthesis, or at least a dialectic, among the various groups studying scientific thinking.
This page was last edited: Wednesday, July 14, 1999