James Gawley’s book examines allusions between poems of different languages with the aim of determining how readers recognize these kinds of intertextual references. The author describes his goal as “a quantitative study of the linguistic features” and a “cognitive study of the mechanics of literary allusion” (p. 9). His test cases are similarities between Vergil’s Aeneid and Homer, and between Voltaire’s epic poem La Henriade (first published 1723) and the Aeneid.
The aims and methods of the study are clearly set out in the Introduction and referred to throughout the book. Chapter 1 attempts to “survey prevalent views on Roman intertextuality and their grounding in the primary source evidence” (p. 15–16). Some positions of Gian Biagio Conte, Donald Andrew Russell and Stephen Hinds on imitatio and aemulatio in Roman literature are briefly discussed. Their works provide the author with a collection of primary source passages, which are then “re-evaluated” according to a clearly defined scheme. To this end, Gawley sorts his findings into four epochs (Punic wars to Early Empire) and four ‘genres’ (poetry, dramatic poetry, oratory, technical prose). He examines which of three types of intertextuality (viz. use of the same topic, imitation of style, direct borrowing) the respective sources comment on and asks whether they reveal positive or negative attitudes towards the discussed type. As a result, he concludes that attitudes towards imitatio changed from the Punic Wars to Vergil. Only in the Augustan period did it appear to be opportune for authors to borrow directly from Latin predecessors, and to refer to a greater extent also to individual phrases from Greek works. After Vergil, who set new standards in terms of the density of allusions, direct borrowing became uncontroversial.
With Chapter 2—in this reviewer’s opinion the most interesting part of the study—the focus shifts from ancient opinions on intertextual practice to modern readers and their recognition of intertextual references. Its goal is to examine a possible statistical correlation between the degree of similarity of a given Vergil-Homer parallel and its interpretation as an allusion by modern test readers. The chapter starts with a criticism of philological evaluations of allusions: according to the author, they are mostly subjective, based only on individual cases, overvalue the criterion of interesting semantic enrichment, and neglect the impact of formal similarity as an argument in favour of or against reading a passage as an allusion (p. 51–55). The author, on the contrary, attempts to test the hypothesis that “form” – similarity between two given passages – can predict “content” – the reader’s interpretation as an allusion (p. 55).
His starting point is Georg Nicolaus Knauer’s monumental study of Vergil’s references to Homer in the Aeneid.[1] Knauer based his work on a corpus of all the Homeric parallels proposed in the history of Vergilian reception which he assembled from Servius and Macrobius to around 1960 and which he provided as an appendix, adding notes on the probability of the respective parallels. From Knauer’s list Gawley chose a selection of 1000 parallels which he presented to four annotators, who were tasked to note correspondences between the proposed Vergilian and Homeric verses in terms of corresponding words, syntax, sound, and topic. The annotators were also asked to rate on a five-point scale how likely they thought it was that the parallel was intended as an allusion by Vergil (cf. Appendix B). In order to determine the degree of formal similarity between a Vergilian and Homeric passage, the individual categories were then converted into numerical values (cf. Appendix A, p. 173–175).
Having ensured by statistical tests that the assessments of the four annotators agreed to a sufficient degree (p. 62–66), Gawley uses a linear regression model to calculate whether the annotators’ ratings on intentionality change depending on the degree of similarity – in other words, whether the degree of similarity can be used to predict the probability with which annotators will rate a parallel as an intended allusion (cf. Appendix B, p. 179, with a link to code and data).[2] As the author points out, the results show that more than 90% of the annotators’ ratings were predictable in this way. Their relevance is confirmed by the fact that these evaluations largely correspond to the indications that Knauer added to his list of parallels in Aeneid 1 regarding their probability as allusions. According to Gawley, this indicates that Knauer’s interpretation should also be predictable on the basis of formal criteria (p. 68–71).
From the combination of formal features, Gawley finally derives a typical pattern for the “clear and resounding” Vergilian allusion: it is on average one verse long, has one linguistic and two topic matches with the targeted Homeric passage, and is attended by up to two further topological reference points that can be found in the Homeric context. Moreover, in the context of the Aeneid it is typically surrounded by a cluster of less clear “generic” parallels (p. 78).
Chapter 3 deals with the unconscious cognitive processes that control memory and attention when recognising allusions. The author compares the list of Vergil-Homer parallels provided by Knauer with a similar list generated by the Tesserae search engine which uses bigrams of matching words in the source and target texts for identifying allusions (p. 82–85). From the remarkably low similarities between the two lists, he concludes that the memory of human readers in the case of interlingual allusions is apparently not simply triggered by lexical similarities. As a second step, he tests the hypothesis that imagistic, “concrete” language is the decisive factor in recognising interlingual allusions (p. 95–96). His analysis of concreteness scores shows that allusive lexis found by human readers is significantly more concrete than non-allusive passages in Vergil, but also than the correspondences found via Tesserae, which do not take concreteness into account (p. 100).
In Chapter 4, Gawley sets out to find out whether the established characteristics of Vergilian allusion – a certain degree of similarity, concrete-figurative language, a mixture of unspecific-“generic” echoes and clear parallels – can be linked to universal cognitive mechanisms (p. 111). He tests this by means of an experimental study in which a test group of 400 readers was presented with various reduced versions of the Cinderella story. Among the conclusions he draws from the results is that the recognition of the Cinderella story does not necessarily occur where the test subjects saw the highest similarity in the test text, but that it was triggered by preceding unspecific “generic” echoes (e.g. “once upon a time”). Furthermore, the higher the concreteness of the text version, the more reliably the Cinderella story was recognised (p. 141–142).
Chapter 5 poses the question of whether the type of “successful allusion” found in the Aeneid is typical only of Vergil’s style or applies to cross-lingual allusions regardless of language and culture. To this end, Gawley repeats the setting of Chapters 2 and 3 in a reduced form for parallels between Voltaire’s Henriade and Vergil’s Aeneid. Here the results of the search with Tesserae are much better, the search engine finding many more parallels, many of them previously unknown. After analysing match words and concreteness, Gawley concludes that Voltaire’s language is less concrete overall in his allusions. Many of Voltaire’s intended allusions would therefore not be recognised by human readers. This confirms the universal nature of the cognitive constraints identified.
In his Conclusions, the author summarises the results of his investigation (p. 165–169). He concludes that the “typical” Vergilian allusion to Homer has specific formal features that ensure its success. It is based not only on lexical similarities, but also on topical correspondences, uses particularly “imagistic” language, and is accompanied by unspecific generic echoes (e.g. type scenes and epic language) that guide the overall process of recognition. Together with the possibility to add “meaning,” such a set of features tends to lead readers to believe that the observed parallel is intended as an allusion. These observations support the assumption of “Dual Coding Theory,” which states that memory is organised primarily via images and the imagination of actions so that allusions to a text in a different language than the source text are possible even without lexical matches. Texts that do not take such cognitive “constraints” into account run the risk that allusions remain “unsuccessful” because they escape the reader’s attention.
The undoubted merits of the work include its lucid methodological reflection, clear argumentation and jargon-free language, which allows even those who are not familiar with statistical methods to follow almost every step of the investigation. Its original approach combines topics and questions from different disciplines – classical studies, digital humanities, cognitive science, Enlightenment studies – in order to analyse a topic with manifold ramifications in a fresh way.
However, it is essentially an empirical and quantitative study, which means that it uses statistical methods, tests hypotheses, and arrives at results only in terms of tendencies and within a certain range of significance. This corresponds both to Gawley’s – from the humanist and classicist point of view somewhat strange – initial question as to whether certain characteristics can “predict” a certain reader reaction, and to his result, the “typical” Vergilian allusion, which in turn confirms a certain hypothesis. As another consequence, the study is about recognising allusions, but not about understanding their meaning, and it is interested in universal cognitive mechanisms, but not in the differences between historical and modern readers and their possibly different reading practises. The fact that Aeneid and La Henriade form the object of investigation seems to a certain extent to be a coincidence; other allusive texts could in principle have led to similar results.
As a reader trained in Classics who has long been interested in the concepts of intertextuality, allusion and quotation held by ancient readers, I find Gawley’s results largely consistent with my own observations, even if I have never thought about a “typical allusion” in terms of significance ranges. I can also accept the methods and results of the study as long as they are intended to answer quantitative questions.
The work becomes problematic, however, when the author ventures into philological territory. This is the case in Chapter 1, where he imitates close reading by categorising “source evidence” compiled from secondary literature into a ready-made scheme in order to finally derive “tendencies,” which in reality are based on only sporadic evidence and remain extremely schematic. Established philological tools (TLL, critical editions) are completely ignored and for the most part also the rich research discussion on theory and practise of imitatio, intertextuality, allusion, quotation and “Quellenkritik.” Another example is Chapter 5, where the idea that Voltaire’s Henriade deliberately avoided figurative language is derived from a single remark on similes and attested only 25 years after the first edition of the poem (p. 144–145). In the same chapter, Gawley sees hidden criticism on the Bourbon kings in the Henriade on the undiscussed assumption that Voltaire advocated a “pessimistic” reading of the Aeneid (p. 149–150) – without any reference to a centuries-long tradition of “affirmative” readings of Vergil’s poem.
The most irritating point, however, is the careless treatment of references throughout the book, citations in the footnotes are inconsistent, incomplete or simply wrong; more than 25 cited titles are missing from the bibliography, including all cited titles of Voltaire’s works; more than 15 journal articles are given without page numbers; names are misspelled; titles by the same author with the same year are not differentiated and the Latin quotations contain numerous typos. As a result, central parts of the argument are impossible to follow, and the credibility and value of the work as a whole are seriously affected.
I come to this conclusion with some regret, because I am deeply convinced that an alignment of hermeneutic and quantitative approaches, of “close reading” and “distant reading” is fruitful and necessary in both directions. This book is an interesting and committed, but in important parts unconvincing attempt.
Notes
[1] Knauer, Georg Nicolaus (1964): Die Aeneis und Homer. Studien zur poetischen Technik Vergils mit Listen der Homerzitate in der Aeneis. Göttingen.
[2] https://github.com/jamesgawley/annotations_of_vergils_allusions (seen 09.07.2025).