Guideline for improving the reliability of Google Ngram studies: Evidence from religious terms

Studying large language models as compression algorithms for human culture

Large language models (LLMs) extract and reproduce the statistical regularities in their training data. Researchers can use these models to study the conceptual relationships encoded in this training data (i.e., the open internet), providing a remarkable opportunity to understand the cultural distinctions embedded within much of recorded human communication.

On the Connection Between Language Change and Language Processing

Previous studies provided evidence for a connection between language processing and language change. We add to these studies with an exploration of the influence of lexical-distributional properties of words in orthographic space, semantic space, and the mapping between orthographic and semantic space on the probability of lexical extinction. Through a binomial linear regression analysis, we investigated the probability of lexical extinction by the first decade of the twenty-first century (2000s) for words that existed in the first decade of the nineteenth-century (1800s) in eight data sets for five languages: English, French, German, Italian, and Spanish. The binomial linear regression analysis revealed that words that are more similar in form to other words are less likely to disappear from a language. By contrast, words that are more similar in meaning to other words are more likely to become extinct. In addition, a more consistent mapping between form and meaning protects a word from lexical extinction. A nonlinear time-to-event analysis furthermore revealed that the position of a word in orthographic and semantic space continues to influence the probability of it disappearing from a language for at least 200 years. Effects of the lexical-distributional properties of words under investigation here have been reported in the language processing literature as well. The results reported here, therefore, fit well with a usage-based approach to language change, which holds that language change is at least to some extent connected to cognitive mechanisms in the human brain.

Multi-LEX: A database of multi-word frequencies for French and English

Written word frequency is a key variable used in many psycholinguistic studies and is central in explaining visual word recognition. Indeed, methodological advances on single-word frequency estimates have helped to uncover novel language-related cognitive processes, fostering new ideas and studies. In an attempt to support and promote research on a related emerging topic, visual multi-word recognition, we extracted from the exhaustive Google Ngram datasets a selection of millions of multi-word sequences and computed their associated frequency estimate. Such sequences are presented with part-of-speech information for each individual word. An online behavioral investigation making use of the French 4-gram lexicon in a grammatical decision task was carried out. The results show an item-level frequency effect of word sequences. Moreover, the proposed datasets were found useful during the stimulus selection phase, allowing more precise control of the multi-word characteristics.

Benford's Law applies to word frequency rank in English, German, French, Spanish, and Italian

Benford's Law states that, in many real-world data sets, the frequency of numbers' first digits is predicted by the formula log(1 + (1/d)). Numbers beginning with a 1 occur roughly 30% of the time, and are six times more common than numbers beginning with a 9. We show that Benford's Law applies to the the frequency rank of words in English, German, French, Spanish, and Italian. We calculated the frequency rank of words in the Google Ngram Viewer corpora. Then, using the first significant digit of the frequency rank, we found the FSD distribution adhered to the expected Benford's Law distribution. Over a series of additional corpora from sources ranging from news to books to social media and across the languages studied, we consistently found adherence to Benford's Law. Furthermore, at the user-level on social media, we found Benford's Law holds for the vast majority of users' collected posts and significant deviations from Benford's Law tends to be a mark of spam bots.

Evaluating the use of Instagram images color histograms and hashtags sets for automatic image annotation

Color similarity has been a key feature for content-based image retrieval by contemporary search engines, such as Google. In this study, we compare the visual content information of images, obtained through color histograms, with their corresponding hashtag sets in the case of Instagram posts. In previous studies, we had concluded that less than 25% of Instagram hashtags are related to the actual visual content of the image they accompany. Thus, the use of Instagram images' corresponding hashtags for automatic image annotation is questionable. In this study, we are answering this question through the computational comparison of images' low-level characteristics with the semantic and syntactic information of their corresponding hashtags. The main conclusion of our study on 26 different subjects (concepts) is that color histograms and filtered hashtag sets, although related, should be better seen as a complementary source for image retrieval and automatic image annotation.

Effect of IUCN Red List category on public attention to mammals

Cultural data is a powerful tool to analyze public awareness of key societal issues, including the conservation of nature. I used two publicly available repositories of cultural data, Google Trends and Google Ngram, to quantify the effect of the International Union for the Conservation of Nature (IUCN) Red List conservation status on public attention toward 4539 mammal species. With Google Trends, I calculated whether Google searches for their common and scientific names have been increasing or decreasing over time. I also ran an anomaly detection analysis to investigate whether a change in red-list status directly results in an increase in Google searches. Additionally, I quantified the mentions of species' common and scientific names in English texts with Google Ngram. Overall, Google searches for most mammal species remained at similar levels or increased since 2008. The severity of species' IUCN Red List status was a significant predictor of increasing Google searches, although the effect size was relatively small. Red-list status seemed strongly confounded with mammal body size. Species that moved to a higher-risk category spiked significantly in Google searches directly after the new designation. The mention of species' common names in the Google Ngram's English 2019 corpus significantly increased as the red-list category increased. These results provide valuable insight into the importance of the IUCN Red List for increasing public awareness and the usefulness of publicly available cultural data on examining the effectiveness of specific conservation efforts and thus evaluating targets for support and funding.

Increasing digitalization is associated with anxiety and depression: A Google Ngram analysis

The prevalence of anxiety disorders and depression are rising worldwide. Studies investigating risk factors on a societal level leading to these rises are so far limited to social-economic status, social capital, and unemployment, while most such studies rely on self-reports to investigate these factors. Therefore, our study aims to evaluate the impact of an additional factor on a societal level, namely digitalization, by using a linguistic big data approach. We extend related work by using the Google Books Ngram Viewer (Google Ngram) to retrieve and adjust word frequencies from a large corpus of books (8 million books or 6 percent of all books ever published) and to subsequently investigate word changes in terms of anxiety disorders, depression, and digitalization. Our analyses comprise and compare data from six languages, British English, German, Spanish, Russian, French, and Italian. We also retrieved word frequencies for the control construct "religion". Our results show an increase in word frequency for anxiety, depression, and digitalization over the last 50 years (r = .79 to .89, p < .001), a significant correlation between the frequency of anxiety and depression words (r = .98, p < .001), a significant correlation between the frequency of anxiety and digitalization words (r = .81, p < .001), and a significant correlation between the frequency of depression and anxiety words (r = .81, p < .001). For the control construct religion, we found no significant correlations for word frequency over the last 50 years and no significant correlation between the frequency of anxiety and depression words. Our results showed a negative correlation between the frequency of depression and religion words (r = -.25, p < .05). We also improved the method by excluding terms with double meanings detected by 73 independent native speakers. Implications for future research and professional and clinical implications of these findings are discussed.

Moral foundations tracked over 200 years of lexicographic data, and their predictors

The prediction that reduction of negative selection decreases group-level competitiveness, as reflected in increased individual-focused and diminished group-focused moral foundations, is tested. To measure this hypothesized shift in moral foundations, we conduct a culturomic analysis of the utilization frequencies of items sourced from the moral foundations item pool, tracked among Britannic populations from 1800 to 1999 using Google Ngram Viewer. The resultant higher-order factor, which tracks increasing individualizing values and decreasing binding values, is termed Asabiyyah (capturing social cohesion and collective purpose). Two predictors of this factor are examined: change in the strength of intergroup competition and change in levels of indicators of developmental instability. Both the strength of intergroup competition and levels of developmental instability associate with Asabiyyah. Rising developmental instability mediates the impact of inter-group competition, indicating that reduced between-group competition might have relaxed negative selection against mutations, which might reduce Asabiyyah via their effects on inter-genomic transactions. These results must be interpreted carefully, given the clear real-world evidence that explicit commitment to group-oriented values often features in harmful and maladaptive social and political ideologies of an extreme character.

Historical representations of social groups across 200 years of word embeddings from Google Books

Using word embeddings from 850 billion words in English-language Google Books, we provide an extensive analysis of historical change and stability in social group representations (stereotypes) across a long timeframe (from 1800 to 1999), for a large number of social group targets (Black, White, Asian, Irish, Hispanic, Native American, Man, Woman, Old, Young, Fat, Thin, Rich, Poor), and their emergent, bottom-up associations with 14,000 words and a subset of 600 traits. The results provide a nuanced picture of change and persistence in stereotypes across 200 y. Change was observed in the top-associated words and traits: Whether analyzing the top 10 or 50 associates, at least 50% of top associates changed across successive decades. Despite this changing content of top-associated words, the average valence (positivity/negativity) of these top stereotypes was generally persistent. Ultimately, through advances in the availability of historical word embeddings, this study offers a comprehensive characterization of both change and persistence in social group representations as revealed through books of the English-speaking world from 1800 to 1999.

The rise and fall of rationality in language

The post-truth era has taken many by surprise. Here, we use massive language analysis to demonstrate that the rise of fact-free argumentation may perhaps be understood as part of a deeper change. After the year 1850, the use of sentiment-laden words in Google Books declined systematically, while the use of words associated with fact-based argumentation rose steadily. This pattern reversed in the 1980s, and this change accelerated around 2007, when across languages, the frequency of fact-related words dropped while emotion-laden language surged, a trend paralleled by a shift from collectivistic to individualistic language.

The surge of post-truth political argumentation suggests that we are living in a special historical period when it comes to the balance between emotion and reasoning. To explore if this is indeed the case, we analyze language in millions of books covering the period from 1850 to 2019 represented in Google nGram data. We show that the use of words associated with rationality, such as “determine” and “conclusion,” rose systematically after 1850, while words related to human experience such as “feel” and “believe” declined. This pattern reversed over the past decades, paralleled by a shift from a collectivistic to an individualistic focus as reflected, among other things, by the ratio of singular to plural pronouns such as “I”/”we” and “he”/”they.” Interpreting this synchronous sea change in book language remains challenging. However, as we show, the nature of this reversal occurs in fiction as well as nonfiction. Moreover, the pattern of change in the ratio between sentiment and rationality flag words since 1850 also occurs in New York Times articles, suggesting that it is not an artifact of the book corpora we analyzed. Finally, we show that word trends in books parallel trends in corresponding Google search terms, supporting the idea that changes in book language do in part reflect changes in interest. All in all, our results suggest that over the past decades, there has been a marked shift in public interest from the collective to the individual, and from rationality toward emotion.

Web-Based Research in Psychology

Abstract. The present article reviews web-based research in psychology. It captures principles, learnings, and trends in several types of web-based research that show similar developments related to web technology and its major shifts (e.g., appearance of search engines, browser wars, deep web, commercialization, web services, HTML5…) as well as distinct challenges. The types of web-based research discussed are web surveys and questionnaire research, web-based tests, web experiments, Mobile Experience Sampling, and non-reactive web research, including big data. A number of web-based methods are presented and discussed that turned out to become important in research methodology. These are one-item-one-screen design, seriousness check, instruction manipulation and other attention checks, multiple site entry technique, subsampling technique, warm-up technique, and web-based measurement. Pitfalls and best practices are described then, especially regarding dropout and other non-response, recruitment of participants, and interaction between technology and psychological factors. The review concludes with a discussion of important concepts that have developed over 25 years and an outlook on future developments in web-based research.

COVID-19 shifts mortality salience, activities, and values in the United States: Big data analysis of online adaptation

What is the effect of a life-threatening pandemic at the societal level? An expanded Theory of Social Change, Cultural Evolution, and Human Development predicts that, during a period of increasing survival threat and decreasing prosperity, humans will shift toward the psychology and behavior typical of the small-scale, collectivistic, and rural subsistence ecologies in which we evolved. In particular, subjective mortality salience, engagement in subsistence activities, and collectivism will all increase, while the aspiration to be wealthy will decrease. Because coronavirus has forced unprecedented proportions of human activity online, we tested hypotheses derived from the theory by analyzing big data samples for 70 days before and 70 days after the coronavirus pandemic stimulated President Trump to declare a national emergency. Google searches were used for an exploratory study; the exploratory study was followed by three independent replications on Twitter, internet forums, and blogs. Across all four internet platforms, terms related to subjective mortality salience, engagement in subsistence activities, and collectivism showed massive increases. These findings, coupled with prior research testing this theory, indicate that humans may have an evolutionarily conditioned response to the level of death and availability of material resources in society. More specifically, humans may shift their behavior and psychology toward that found in subsistence ecologies under conditions of high mortality and low prosperity or, conversely, toward behavior and psychology found in modern commercial ecologies under conditions of low mortality and high prosperity.

Key issues in professionalizing mentoring practices

Mentoring has experienced a tremendous upswing over the past decades, which has only recently slowed down somewhat. One possible factor explaining mentoring's popularity are numerous case studies suggesting that it is one of the most effective ways of helping individuals to develop. Meta-analyses indicating effect sizes for mentoring that are below what would theoretically be possible appear to contradict the success stories, however. This circumstance raises questions about the professionalization of mentoring practices. We focus on seven key issues for future efforts at professionalizing mentoring. Key issues 1 and 2 address observation of the state of the art within formal mentoring when programs are planned and implemented: the consideration of recent research and of best practices. While both areas can overlap, they provide complementary sources of pertinent information for the professionalization of mentoring. Key issues 3-6 address the need to align mentoring activities to the specific context and goals of individual mentoring programs by observing idiographic program characteristics, mentoring dynamics, the orchestration of mentoring goals, and the provision of mentoring resources. Finally, key issue 7 highlights ongoing evaluation as the basis of the effective, continuous improvement of mentoring programs.

A sensorimotor control framework for understanding emotional communication and regulation

Our research team was asked to consider the relationship of the neuroscience of sensorimotor control to the language of emotions and feelings. Actions are the principal means for the communication of emotions and feelings in both humans and other animals, and the allostatic mechanisms controlling action also apply to the regulation of emotional states by the self and others. We consider how motor control of hierarchically organised, feedback-based, goal-directed action has evolved in humans, within a context of consciousness, appraisal and cultural learning, to serve emotions and feelings. In our linguistic analysis, we found that many emotion and feelings words could be assigned to stages in the sensorimotor learning process, but the assignment was often arbitrary. The embodied nature of emotional communication means that action words are frequently used, but that the meanings or senses of the word depend on its contextual use, just as the relationship of an action to an emotion is also contextually dependent.

Anticipatory feelings: Neural correlates and linguistic markers

This review introduces anticipatory feelings (AF) as a new construct related to the process of anticipation and prediction of future events. AF, defined as the state of awareness of physiological and neurocognitive changes that occur within an oganism in the form of a process of adapting to future events, are an important component of anticipation and expectancy. They encompass bodily-related interoceptive and affective components and are influenced by intrapersonal and dispositional factors, such as optimism, hope, pessimism, or worry. In the present review, we consider evidence from animal and human research, including neuroimaging studies, to characterize the brain structures and brain networks involved in AF. The majority of studies reviewed revealed three brain regions involved in future oriented feelings: 1) the insula; 2) the ventromedial prefrontal cortex (vmPFC); and 3) the amygdala. Moreover, these brain regions were confirmed by a meta-analysis, using a platform for large-scale, automated synthesis of fMRI data. Finally, by adopting a neurolinguistic and a big data approach, we illustrate how AF are expressed in language.