Language entails many nested time scales, ranging from therelatively slow scale of cultural evolution to the rapid scale ofindividual cognition. The nested, multiscale nature oflanguage implies that even simple acts of text production,such as typing a sentence, entail complex interactionsinvolving multiple concurrent processes. As such, textproduction may have much in common with other cognitivephenomena thought to emerge from multiplicativeinteractions across temporal scales, namely those that exhibitfractal properties. We investigated the relationship betweenfractal scaling and the quality of produced text. Participants(N=131) wrote essays while their keystrokes were recorded.Fractal analyses were then performed on time series ofinterkeystroke intervals (IKIs). Results showed that fractalproperties characterizing IKIs positively predicted expertratings of essay quality, even after accounting for essaylength. The results support our hypotheses concerningmultiscale coordination and text production.

The relationship between working memory capacity andwriting ability was examined via a linguistic analysis ofstudent essays. Undergraduate students (n = 108) wrotetimed, prompt-based essays and completed a battery ofcognitive assessments. The surface- and discourse-levellinguistic features of students’ essays were then analyzedusing natural language processing tools. The results indicatedthat WM capacity was related to surface-level, but notdiscourse-level features of student essays. Additionally, theresults suggest that these relationships were attenuated forstudents with high inferencing skills, as opposed to those withlower inferencing skills

This study investigates how cohesion manifests in readers’thought processes while reading texts when they areinstructed to engage in self-explanation, a strategy associatedwith deeper, more successful comprehension. In Study 1,college students (n = 21) were instructed to either paraphraseor self-explain science texts. Paraphrasing was characterizedby greater cohesion in terms of lexical overlap whereas self-explanation included greater lexical diversity and moreconnectives to specify relations between ideas. In Study 2,adolescent students (n = 84) were provided with instructionand practice in self-explanation and reading strategies across8 sessions. Self-explanations increased in lexical diversity butbecame more causally and semantically cohesive over time.Together, these results suggest that cohesive featuresexpressed in think alouds are indicative of the depth ofstudents’ comprehension processes.

We investigated the differential effects of self-explaining a refutational text, compared to thinking aloud or rereading. Undergraduate students (n = 105) read a refutational text about natural selection and were asked to either self-explain, think-aloud, or re-read the text. Then they completed a posttest that assessed general knowledge of natural selection. Students who self-explained the refutational text subsequently outperformed their peers on a test of their knowledge of natural selection. Additionally, the results suggest that both instructional and performance differences were significantly linked to the degree of causal cohesion present within students’ natural language responses to the text (i.e., selfexplanations and think-alouds).

This study investigates the potential for linguistic microfeatures related to length, complexity, cohesion, relevance, topic, and rhetorical style to predict L2 writing proficiency. Computational indices were calculated by two automated text analysis tools (Coh-Metrix and the Writing Assessment Tool) and used to predict human essay ratings in a corpus of 480 independent essays written for the TOEFL. A stepwise regression analysis indicated that six linguistic microfeatures explained 60% of the variance in human scores for essays in a test set, providing an exact accuracy of 55% and an adjacent accuracy of 96%. To examine the limitations of the model, a post-hoc analysis was conducted to investigate differences in the scoring outcomes produced by the model and the human raters for essays with score differences of two or greater (N = 20). Essays scored as high by the regression model and low by human raters contained more word types and perfect tense forms compared to essays scored high by humans and low by the regression model. Essays scored high by humans but low by the regression model had greater coherence, syntactic variety, syntactic accuracy, word choices, idiomaticity, vocabulary range, and spelling accuracy as compared to essays scored high by the model but low by humans. Overall, findings from this study provide important information about how linguistic microfeatures can predict L2 essay quality for TOEFL-type exams and about the strengths and weaknesses of automatic essay scoring models.

Text cohesion is an important element of discourseprocessing. This paper presents a new approach to modeling,quantifying, and visualizing text cohesion using automatedcohesion flow indices that capture semantic links amongparagraphs. Cohesion flow is calculated by applyingCohesion Network Analysis, a combination of semanticdistances, Latent Semantic Analysis, and Latent DirichletAllocation, as well as Social Network Analysis. Experimentsperformed on 315 timed essays indicated that cohesion flowindices are significantly correlated with human ratings of textcoherence and essay quality. Visualizations of the globalcohesion indices are also included to support a more facileunderstanding of how cohesion flow impacts coherence interms of semantic dependencies between paragraphs.

Dialogism provides the grounds for building a comprehensivemodel of discourse and it is focused on the multiplicity ofperspectives (i.e., voices). Dialogism can be present in anytype of text, while voices become themes or recurrent topicsemerging from the discourse. In this study, we examine theextent that differences between self-explanations and think-alouds can be detected using computational textual indicesderived from dialogism. Students (n = 68) read a text aboutnatural selection and were instructed to generate self-explanations or think-alouds. The linguistic features of thesetext responses were analyzed using ReaderBench, anautomated text analysis tool. A discriminant function analysisusing these features correctly classified 80.9% of the students’assigned experimental conditions (self-explanation vs. thinkaloud). Our results indicate that self-explanation promotestext processing that focuses on connected ideas, rather thanseparate voices or points of view covering multiple topics.