Lexical items in the vocabulary of a language undergo dramatic changes over time, explaining the mechanismsthat cause this change has been an important topic for the cognitive sciences. One particular focus for researchers has beenunderstanding the dynamics of change in word forms. The rate (or half-life) at which word forms change over time variesgreatly, and corpus-based cladistic studies have shown that certain properties, such as word frequency, length and age ofacquisition, can be used to predict this variation. We test through the use of an artificial language learning paradigm the extentto which these psycholinguistic factors affect accurate learning of word forms, linking processes of acquisition with processesof evolutionary change. Our findings provide an insight into the underlying mechanisms that drive diachronic change within alanguage’s vocabulary, highlighting the important role that the learning process has on lexical change.

Quantifying the meaning of a word is a complex challenge. Humans can encode semantic information along a large and diverse range of semantic dimensions for any given word. Whilst a number of studies have applied a range of techniques to quantify word meaning along specific dimensions, little work has focussed on the socio-semantic dimensions of meaning. Here, we present data that quantifies the socio-semantic representations of 2,700 Czech words along the dimensions of gender, location, political, valence and age. We also demonstrate the utility of the data set by calculating an estimate of socio-semantic similarity between all words, which can be used to identify words that are either proximally close or distant in socio-semantic space.

Semantic representations of words can vary from person to person and can undergo change over the lifespan. Although a number of studies have explored differences in affective variables (e.g. valence), there has not been any large-scale investigation of the way socio-semantic information is represented, nor whether there is variation (or stability) in these representations across a diverse participant sample. We report on a study where 2,700 Czech words were rated by 1,450 Czech speakers along the dimensions of gender (feminine/masculine), location (rural/urban), political alignment (conservative/liberal), valence (negative/positive) and age (old/young). The ratings were analysed based on gender and age demographics. Our results demonstrate that distinct categories of words, e.g. those associated with mental wellbeing (such as úzkost [anxiety]) are self-associated – young participants rate the words as more young, older participants as more old. Our results shed light on the variation that exists in socio-semantic representations within different populations.

Learning a language involves learning how to map specificforms onto their associated meanings. Such mappings canutilise arbitrariness and non-arbitrariness, yet, ourunderstanding of how these two systems operate at differentstages of vocabulary development is still not fully understood.The Sound-Symbolism Bootstrapping Hypothesis (SSBH)proposes that sound-symbolism is essential for word learningto commence, but empirical evidence of exactly how sound-symbolism influences language learning is still sparse. It maybe the case that sound-symbolism supports acquisition ofcategories of meaning, or that it enables acquisition ofindividualized word meanings. In two Experiments whereparticipants learned form-meaning mappings from eithersound-symbolic or arbitrary languages, we demonstrate thechanging roles of sound-symbolism and arbitrariness fordifferent vocabulary sizes, showing that sound-symbolismprovides an advantage for learning of broad categories, whichmay then transfer to support learning individual words,whereas an arbitrary language impedes acquisition ofcategories of sound to meaning.

In two behavioral experiments, we demonstrated that human observers can extract average body size from a group ofindividuals. In Experiment 1, we asked 38 participants to estimate the average body size from a group of 5, 10 or 15bodies that were presented in various angles of view (Profile, Three-Quarter, Frontal, and Mixed). Participants were ableto extract the average body size, but they systematically overestimated thinner body groups, and underestimated largerbody groups. Biases were generally reduced for smaller sets sizes and when bodies were shown in profile view, but thetrend was reversed for sets with larger bodies. In Experiment 2, we tested 37 participants and showed that the accuracyof their estimates was modulated by presentation time: Accuracy was poorest when groups were presented for 1s, butsignificantly improved for 3s and 5s presentations. Implications of these finding are discussed.

Linguistic corpora have been a vital resource for understanding not only how we use language, but also how we process words and sentences. In order to better understand language processing, researchers have recently been creating corpora that integrate both traditional text annotations as well as behavioural measurements collected from human participants. In this paper we introduce the HeCz Corpus, which to our knowledge is the largest such example of a behavioural corpus, containing 1,919 newspaper headlines taken from a Czech language news website. The sample consisted of 1,872 participants, each reading approximately 120 headlines. Each headline was read using a self-paced reading, meaning that every word in the corpus can be analyzed for reading time. After reading each headline, each participant answered a question relating to a specific information contained within the headline, providing a measurement of comprehension. To facilitate better understanding of participant level variation in how the headlines are processed, we collected data on the participant's mood state immediately prior to their participation, along with other basic demographic information. We also collected data from a subset of participants who read the stimuli in the initial testing round, but also completed the same experiment in a second round after a one-month gap, which can provide new insights into how texts are processed and understood when being re-read. In order to highlight the practical uses of the corpus, our analyses focus on how reading times are modulated by i) headline length in words, ii) trial order, and iii) testing round, in addition to examining the role of targeted information location in comprehension accuracy. HeCz thus provides a unique and novel resource that can be used by psycholinguists and cognitive scientists more generally, in order to gain new insights into how real-world language is processed and understood.

Number agreement attraction in comprehension is a phenomenon that has been documented in various typologically diverse languages. This evidence has led to claims about the cross-linguistic uniformity of agreement attraction effects and its independence from the grammatical features of a particular language. However, recent research has shown that in Czech, number agreement attraction effects are either absent or negligible in size. This directly contradicts the cross-linguistic uniformity hypothesis. The current paper aims to further corroborate this finding and presents a direct experimental comparison of Czech and English. Two comparable self-paced reading experiments were conducted using stimuli that were translation equivalents in Czech and English. Our analyses demonstrate a preference for the null model in Czech (no agreement attraction), unlike in English, where an interaction between verb number and attractor number was preferred. Moreover, when we compare the data from the two experiments directly, we find that the interaction between language and attraction was also preferred. In sum, we provide evidence against the cross-linguistic uniformity hypothesis for agreement attraction effects and suggest that processing patterns may differ between languages even for almost identical structures, such as agreement relations.

Many theories of semantic memory assume that categories spontaneously emerge from commonalities in the way we per-ceive and interact with the world around us. However, efforts to test this assumption computationally have been hamperedby use of abstracted features without clear sensorimotor grounding and over-reliance on small samples of concepts from alimited number of categories. Taking a radically different approach, we examined whether categorical structure emergesspontaneously from the latent structure of sensorimotor experience by creating a fully-grounded multidimensional senso-rimotor space at the scale of a full-size human conceptual system (i.e., 11 sensorimotor dimensions x 40,000 concepts).We found evidence for (a) a high-level separation of abstract and concrete categories (which was not enhanced by theinclusion of affective information); (b) a hierarchical structure of concrete concepts that separated categories commonlyimpaired in double dissociations, such as fruit/vegetables, animals, tools, and musical instruments; and (c) a flatter hi-erarchy of abstract concepts that separated categories such as negative emotions, units of time, social relationships, andpolitical systems. These findings demonstrate that grounded sensorimotor information is fundamental to the representationof all conceptual knowledge.

Sensorimotor information plays a fundamental role incognition. However, datasets of ratings of sensorimotorexperience have generally been restricted to several hundredwords, leading to limited linguistic coverage and reducedstatistical power for more complex analyses. Here, we presentmodality-specific and effector-specific norms for 39,954concepts across six sensory modalities (touch, hearing, smell,taste, vision, and interoception) and five action effectors(mouth/throat, hand/arm, foot/leg, head excluding mouth, andtorso), which were gathered from 4,557 participants whocompleted a total of 32,456 surveys using Amazon'sMechanical Turk platform. The dataset therefore representsone of the largest set of semantic norms currently available.We describe the data collection procedures, provide summarydescriptives of the data set, demonstrate the utility of thenorms in predicting lexical decision times and accuracy, aswell as offering new insights and outlining avenues for futureresearch. Our findings will be of interest to researchers inembodied cognition, cognitive semantics, sensorimotorprocessing, and the psychology of language generally. Thescale of this dataset will also facilitate computationalmodelling and big data approaches to the analysis of languageand conceptual representations.

Learning how words refer to aspects of the environment is acomplex task, but one that is supported by numerous cueswithin the environment which constrain the possibilities formatching words to their intended referents. In this paper wetested the predictions of a computational model of multiplecue integration for word learning, that predicted variation inthe presence of cues provides an optimal learning situation. Ina cross-situational learning task with adult participants, wevaried the reliability of presence of distributional, prosodic,and gestural cues. We found that the best learning occurredwhen cues were often present, but not always. The effect ofvariability increased the salience of individual cues for thelearner, but resulted in robust learning that was not vulnerableto individual cues’ presence or absence. Thus, variability ofmultiple cues in the language-learning environment providedthe optimal circumstances for word learning.