Jon W. Carr’s research while affiliated with International School for Advanced Studies and other places

Languages are teeming with statistical cues for the cognitive system to capitalize on, and at least some aspects of language processing might be explained by sensitivity to such cues via general-purpose cognitive machinery. One cue relevant to the task of reading is the way in which information about word identity is typically distributed within a given lexicon. Previous research on visual word recognition has struggled to disentangle information distribution from perceptual factors because they are unavoidably interconnected in the processing of language. We construct a formal model that explicitly teases the two apart and use a novel experimental paradigm with artificial lexicons to test whether humans are sensitive to this cue. Our results provide a causal demonstration that the way in which the lexicon distributes information affects how readers visually explore words, thus revealing the use of another probabilistic cue that is based simply on efficient information gathering.

A common problem in eye-tracking research is vertical drift—the progressive displacement of fixation registrations on the vertical axis that results from a gradual loss of eye-tracker calibration over time. This is particularly problematic in experiments that involve the reading of multiline passages, where it is critical that fixations on one line are not erroneously recorded on an adjacent line. Correction is often performed manually by the researcher, but this process is tedious, time-consuming, and prone to error and inconsistency. Various methods have previously been proposed for the automated, post hoc correction of vertical drift in reading data, but these methods vary greatly, not just in terms of the algorithmic principles on which they are based, but also in terms of their availability, documentation, implementation languages, and so forth. Furthermore, these methods have largely been developed in isolation with little attempt to systematically evaluate them, meaning that drift correction techniques are moving forward blindly. We document ten major algorithms, including two that are novel to this paper, and evaluate them using both simulated and natural eye-tracking data. Our results suggest that a method based on dynamic time warping offers great promise, but we also find that some algorithms are better suited than others to particular types of drift phenomena and reading behavior, allowing us to offer evidence-based advice on algorithm selection.

Experimental and cross-linguistic studies have shown that vocal iconicity is prevalent in words that carry meanings related to size and shape. Although these studies demonstrate the importance of vocal iconicity and reveal the cognitive biases underpinning it, there is less work demonstrating how these biases lead to the evolution of a sound symbolic lexicon in the first place. In this study, we show how words can be shaped by cognitive biases through cultural evolution. Using a simple experimental setup resembling the game telephone, we examined how a single word form changed as it was passed from one participant to the next by a process of immediate iterated learning. About 1,500 naïve participants were recruited online and divided into five condition groups. The participants in the control-group received no information about the meaning of the word they were about to hear, while the participants in the remaining four groups were informed that the word meant either big or small (with the meaning being presented in text), or round or pointy (with the meaning being presented as a picture). The first participant in a transmission chain was presented with a phonetically diverse word and asked to repeat it. Thereafter, the recording of the repeated word was played for the next participant in the same chain. The sounds of the audio recordings were then transcribed and categorized according to six binary sound parameters. By modelling the proportion of vowels or consonants for each sound parameter, the small-condition showed increases of front unrounded vowels and the pointy-condition increases of acute consonants. The results show that linguistic transmission is sufficient for vocal iconicity to emerge, which demonstrates the role non-arbitrary associations play in the evolution of language.

Recent research has shown that semantic category systems, such as color and kinship terms, find an optimal balance between simplicity and informativeness. We argue that this situation arises through pressure for simplicity from learning and pressure for informativeness from communicative interaction, two distinct pressures that often (but not always) pull in opposite directions. Another account argues that learning might also act as a pressure for informativeness, that learners might be biased toward inferring informative systems. This results in two competing hypotheses about the human inductive bias. We formalize these competing hypotheses in a Bayesian iterated learning model in order to simulate what kinds of languages are expected to emerge under each. We then test this model experimentally to investigate whether learners' biases, isolated from any communicative task, are better characterized as favoring simplicity or informativeness. We find strong evidence to support the simplicity account. Furthermore, we show how the application of a simplicity principle in learning can give the impression of a bias for informativeness, even when no such bias is present. Our findings suggest that semantic categories are learned through domain-general principles, negating the need to posit a domain-specific mechanism.

A common problem in eye tracking research is vertical drift—the progressive displacement of fixation registrations on the vertical axis that results from a gradual loss of eye tracker calibration over time. This is particularly problematic in experiments that involve the reading of multiline passages, where it is critical that fixations on one line are not erroneously recorded on an adjacent line. Correction is often performed manually by the researcher, but this process is tedious, time-consuming, and prone to error and inconsistency. Various methods have previously been proposed for the automated, post-hoc correction of vertical drift in reading data, but these methods vary greatly, not just in terms of the algorithmic principles on which they are based, but also in terms of their availability, documentation, implementation languages, and so forth. Furthermore, these methods have largely been developed in isolation with little attempt to systematically evaluate them, meaning that drift correction techniques are moving forward blindly. We document nine major algorithms, including two that are novel to this paper, and evaluate them using both simulated and natural eye tracking data. Our results suggest that a method based on dynamic time warping offers great promise, but we also find that some algorithms are better suited than others to particular types of drift phenomena and reading behavior, allowing us to offer evidence-based advice on algorithm selection.

Recent research has shown that semantic category systems, such as color and kinship terms, find an optimal balance between the considerations of simplicity and informativeness. We argue that this situation arises through a pressure for simplicity from learning and a pressure for informativeness from communicative interaction, two distinct pressures that pull in (often but not always) opposite directions. An alternative account suggests that learning might also act as a pressure for informativeness—that learners might be biased toward inferring informative systems. This results in two competing hypotheses about the human inductive bias. We formalize these competing hypotheses in a Bayesian iterated learning model and test them in two experiments with human participants. Specifically, we investigate whether learners' inductive biases, isolated from any communicative task, are better characterized as favoring simplicity or informativeness. We find strong evidence to support the simplicity account. Furthermore, we show how the application of a simplicity principle in learning can give the impression of a bias for informativeness, even when no such bias is present. Our findings suggest that semantic categories are learned through domain-general principles, negating the need to posit a domain-specific inductive bias.

Appendix S1. Experimental briefs