Moretti begins with "Graphs". One example charts the number of new novels published each year, in Japan, Denmark, France and Italy, and attempts to connect these to political and social changes. (He segues from here to a chart of the value of book imports into colonial India, without mentioning that this is a rather different metric.) Another considers long-term changes in the gender breakdown of authorship of British novels.
The core case study, however, is of genre in British novels. A chart of publication numbers nicely illustrates the broad rise and fall of epistolary, gothic and historical novels, succeeding one another as the most popular genre from 1750 to 1850. Moretti follows this up by looking at 44 finer-grained novel genres between 1740 and 1900, suggesting these appear in clusters rather than randomly. (He doesn't formally test this, but the statistician Cosma Shalizi has, finding that random genre creation would only produce something resembling the observed pattern with 0.4% probability.) Clustering, however, does not imply cyclical structure and the evidence is much weaker for Moretti's further argument, in which Kuhnian-style revolutions produce regular genre changes: "Normal literature remains in place for twenty-five years or so..." The "or so" here is broad enough to make speculation about "hidden rhythms" and "a note of perplexity" about generational effects quite unnecessary. (It doesn't help Moretti's case here that he excludes detective fiction and science fiction from the analysis because of their "peculiar long duration".)
Turning to "Maps", Moretti looks at Mary Mitford's Our Village, John Galt's Annals of the Parish, and Berthold Auerbach's Black Forest Village Stories. He uses concentric circles to group links from the central village to neighbouring villages, nearby towns, bigger cities, and overseas. These patterns are then connected to increasing state power, central place theory, and the changes brought to English parish geographies by enclosure.
Now the perception of geography can certainly differ from the reality, but even so, at this point it seems to me that we are at a remove from literary history. Whether in villages or cities, people's social and economic lives depend in important ways on the local geography and on connections further afield. And novels about these people reflect these structures and their changes over time. To take a very simple case from Our Village, anyone doing a series of recreational walks from one place is likely to venture in different directions for variety, but to walk roughly the same distance each time, thus generating a circular pattern of destinations... Is any narrative or literary device actually involved here? Meanwhile the more distant links in Moretti's examples don't seem essentially geographical — or even "geometrical", which is the alternative he suggests.
Maps can be a useful way to illustrate aspects of novels or stories; and they are certainly important tools in economic and social history. Whether they "add anything to our knowledge of literature" more directly is not so clear.
The misuse, if not abuse, of evolutionary ideas in other areas has a long and unillustrious history. Moretti is not a particularly egregious offender here, but in "Trees" he appeals to a broad selection of ideas from evolutionary biology and achieves an equally broad range of confusions.
Latching onto the one tree diagram in The Origin of Species, Moretti begins by confusing the idea of "character displacement" with evolution more generally. He suggests that "the basic mechanism of change is divergence", but that is just to relabel the result. The actual mechanisms of evolution are variation and selection, genetic drift, and so forth, which Moretti barely hints at possible literary analogues to. Concepts such as "allopatric speciation" and "morphological novelty" do get an airing, however, along with the difference between population and typological thinking.
The first example involves a bifurcating tree of detective stories by Conan Doyle and contemporaries, with each division involving a classification by the presence or absence of clues, necessary clues, visible clues, and potentially decodable clues. But this is not an evolutionary tree — there is no concept of time — but rather a dichotomous classification key.
He expands this to include dates, presenting a diagram that could be viewed as an evolutionary tree of plot ideas. He also gives us a tree of practitioners of "free indirect style in modern narrative". Though he recognises that both of these should be reticulated networks rather than trees, however, he doesn't follow the implications of that. So a detective story with "decodable" clues may be copying the decodability from the preceding stories in the same lineage, but may also have been been influenced into having clues at all by "visible but not decodable" stories on a sister lineage. And inheritance can skip "generations": Mann is likely to have taken elements of style directly from Goethe, not just via Flaubert. Here biological analogies suggest inappropriate inheritance models and risk obscuring important connections.
Graphs, Maps, Trees presents a lot of ideas without making much of an attempt at testing them: it gives us a lot of variation but not much selection.
Graphs and maps are certainly a useful and suggestive way of presenting information about genres and plots and so forth — and literary history should make full use of informational graphics — but it's not clear that the examples Moretti gives actually provide any analytical depth. They might make nice illustrations for broader studies, but by themselves they are a bit anaemic.
Humans find patterns all too easily, and this extends to the interpretation of graphs, plots, and data sets. Which is why we have developed formal statistical analysis. So students of literary history seeking to follow Moretti in using abstract models should consider finding a statistician to collaborate with, even if the result is just a brief footnote about significance.
Anyone working in the humanities should avoid using terminology from evolutionary biology: nine times out of ten it will be misleading and the tenth time it will be unnecessary. Those evolutionary concepts that are generalisable have already been abstracted for use in other disciplines, and these abstract methods can be used without importing baggage from biology. Here again, intuition is an unreliable guide for outsiders — and biologists have done their share of misapplying evolutionary ideas.
Perhaps all of this is only meant as metaphor, as an afterword by biologist Alberto Piazza half suggests: "literary writing... is capable of metabolizing metaphors and ambiguities belonging to several systems of knowledge". But I think that is to go the wrong way: literary history is just as valid a domain for the application of classification keys or replicator dynamics or network theory as biology or geography or any other field. The question in any particular case is whether such tools are applicable and how useful they are, not whether we can find tenuous parallels with their applications in evolutionary biology or linguistics.
These disagreements notwithstanding, Graphs, Maps, Trees is provocative, entertaining, and concise. A book that can get from lovers in Parisian novels on one page to D'Arcy Thompson's On Growth and Form on the next, that can reference Mikhail Bakhtin alongside Ernst Mayr, must spark something in anyone with any curiosity. Moretti is entertaining, at times even funny — he can comment, for example, that "geography is a useful tool, yes, but does not explain everything. For that, we have astrology and 'Theory'" — and he doesn't take himself too seriously. And the whole book, many full-page graphs and maps included, is only 120 pages long. It is a pleasure to disagree with.