Undercompression errors as evidence for conceptual primitives.

The Meaning First Approach offers a model of the relation between thought and language that includes a Generator and a Compressor. The Generator build non-linguistic thought structures and the Compressor is responsible for its articulation through three processes: structure-preserving linearization, lexification, and compression via non-articulation of concepts when licensed. One goal of this paper is to show that a range of phenomena in child language can be explained in a unified way within the Meaning First Approach by the assumption that children differ from adults with respect to compression and, specifically, that they may undercompress in production, an idea that sets a research agenda for the study of language acquisition. We focus on dependencies involving pronouns or gaps in relative clauses and wh-questions, multi-argument verbal concepts, and antonymic concepts involving negation or other opposites. We present extant evidence from the literature that children produce undercompression errors (a type of commission errors) that are predicted by the Meaning First Approach. We also summarize data that children's comprehension ability provides evidence for the Meaning First Approach prediction that decompression should be challenging, when there is no 1-to-1 correspondence.

Situating language register across the ages, languages, modalities, and cultural aspects: Evidence from complementary methods.

In the present review paper by members of the collaborative research center "Register: Language Users' Knowledge of Situational-Functional Variation" (CRC 1412), we assess the pervasiveness of register phenomena across different time periods, languages, modalities, and cultures. We define "register" as recurring variation in language use depending on the function of language and on the social situation. Informed by rich data, we aim to better understand and model the knowledge involved in situation- and function-based use of language register. In order to achieve this goal, we are using complementary methods and measures. In the review, we start by clarifying the concept of "register", by reviewing the state of the art, and by setting out our methods and modeling goals. Against this background, we discuss three key challenges, two at the methodological level and one at the theoretical level: (1) To better uncover registers in text and spoken corpora, we propose changes to established analytical approaches. (2) To tease apart between-subject variability from the linguistic variability at issue (intra-individual situation-based register variability), we use within-subject designs and the modeling of individuals' social, language, and educational background. (3) We highlight a gap in cognitive modeling, viz. modeling the mental representations of register (processing), and present our first attempts at filling this gap. We argue that the targeted use of multiple complementary methods and measures supports investigating the pervasiveness of register phenomena and yields comprehensive insights into the cross-methodological robustness of register-related language variability. These comprehensive insights in turn provide a solid foundation for associated cognitive modeling.

Meaning and Use in the Expression of Estimative Probability.

Words of estimative probability (WEPs), such as 'possible' and 'a good chance', provide an efficient means for expressing probability under uncertainty. Current semantic theories assume that WEPs denote crisp thresholds on the probability scale, but experimental data indicate that their use is characterised by gradience and focality. Here, we implement and compare computational models of the use of WEPs to explain novel production data. We find that, among models incorporating cognitive limitations and assumptions about goal-directed speech, a model that implements a threshold-based semantics explains the data equally well as a model that semantically encodes patterns of gradience and focality. We further validate the model by distinguishing between participants with more or fewer autistic traits, as measured with the Autism Spectrum Quotient test. These traits include communicative difficulties. We show that these difficulties are reflected in the rationality parameter of the model, which modulates the probability that the speaker selects the pragmatically optimal message.

Probabilistic pragmatics explains gradience and focality in natural language quantification.

An influential view in philosophy and linguistics equates the meaning of a sentence to the conditions under which it is true. But it has been argued that this truth-conditional view is too rigid and that meaning is inherently gradient and revolves around prototypes. Neither of these abstract semantic theories makes direct predictions about quantitative aspects of language use. Hence, we compare these semantic theories empirically by applying probabilistic pragmatic models as a link function connecting linguistic meaning and language use. We consider the use of quantity words (e.g., "some," "all"), which are fundamental to human language and thought. Data from a large-scale production study suggest that quantity words are understood via prototypes. We formulate and compare computational models based on the two views on linguistic meaning. These models also take into account cognitive factors, such as salience and numerosity representation. Statistical and empirical model comparison show that the truth-conditional model explains the production data just as well as the prototype-based model, when the semantics are complemented by a pragmatic module that encodes probabilistic reasoning about the listener's uptake.

Generative Grammar: A Meaning First Approach.

The theory of language must predict the possible thought-signal (or meaning-sound or sign) pairings of a language. We argue for a Meaning First architecture of language where a thought structure is generated first. The thought structure is then realized using language to communicate the thought, to memorize it, or perhaps with another purpose. Our view contrasts with the T-model architecture of mainstream generative grammar, according to which distinct phrase-structural representations-Phonetic Form (PF) for articulation, Logical Form (LF) for interpretation-are generated within the grammar. At the same time, our view differs from early transformational grammar and generative semantics: We view the relationship between the thought structure and the corresponding signal as one of compression. We specify a formal sketch of compression as a choice between multiple possible pronounciations balancing the desire to transmit information against the effort of pronounciation. The Meaning First architecture allows a greater degree of independence between thought structures and the linguistic signal. We present three arguments favoring this type of independence. First we argue that scopal properties can be better explained if we only compare thought structures independent of the their realization as a sentence. Secondly, we argue that Meaning First architecture allows contentful late insertion, an idea that has been argued for in Distributed Morphology already, but as we argue is also motivated by the division of the logical and socio-emotive meaning content of language. Finally, we show that only the Meaning First architecture provides a satisfying account of the mixing of multiple languages by multilingual speakers, especially for cases of simultaneous articulation across two modalities in bimodal speakers. Our view of the structure of grammar leads to a reassessment of priorities in linguistic analyses: while current mainstream work is often focused on establishing one-to-one relationships between concepts and morphemes, our view makes it plausible that primitive concepts are frequently marked indirectly or unpronounced entirely. Our view therefore assigns great value to the understanding of logical primitives and of compression.

Study of perceptions of the extent to which guideline recommendations are binding: a survey of commonly used terminology.

BACKGROUND: Doctors are more likely to implement guidelines in their everyday practice if the recommendations contained in them are understandable. So far, there has been little standardization in the wording of guideline recommendations. It would be important to know how certain terms are understood by guideline users. In this study, doctors were asked in a survey about what they considered to be the level of obligation carried by various formulations that are commonly used in guidelines to recommend particular courses of action. METHODS: An online survey of physicians (mostly dermatologists) was carried out in which they were asked to rate, on a visual analog scale, what they perceived to be the level of obligation of various common formulations for guideline recommendations. RESULTS: The terms "muss" (must) and "darf nicht" (must not) were interpreted as being maximally binding. The two closely related German words "soll" (shall) and "sollte" (should) were considered highly binding, as were negative formulations such as "wird nicht empfohlen" (is not recommended). The perceived level of obligation of "soll" did not differ from that of "sollte" to any detectable extent, nor was there any detectable distinction between the various negative formulations studied. Formulations with the words "wird empfohlen" (is recommended), "kann empfohlen werden" (can be recommended), or other "kann" (can) expressions were considered to be only mildly or moderately binding. In general, there was marked variation in the perceived level of obligation of formulations located in the low and middle ranges. CONCLUSION: These findings suggest that guideline users do not necessarily perceive recommendation strengths as the guideline authors intended. It might be better if positive recommendations came in only two different strengths, while a single recommendation strength might suffice for negative ones. Further studies should shed more light on this question.

Familial sinistrals avoid exact numbers.

We report data from an internet questionnaire of sixty number trivia. Participants were asked for the number of cups in their house, the number of cities they know and 58 other quantities. We compare the answers of familial sinistrals--individuals who are left-handed themselves or have a left-handed close blood-relative--with those of pure familial dextrals--right-handed individuals who reported only having right-handed close blood-relatives. We show that familial sinistrals use rounder numbers than pure familial dextrals in the survey responses. Round numbers in the decimal system are those that are multiples of powers of 10 or of half or a quarter of a power of 10. Roundness is a gradient concept, e.g. 100 is rounder than 50 or 200. We show that very round number like 100 and 1000 are used with 25% greater likelihood by familial sinistrals than by pure familial dextrals, while pure familial dextrals are more likely to use less round numbers such as 25, 60, and 200. We then use Sigurd's (1988, Language in Society) index of the roundness of a number and report that familial sinistrals' responses are significantly rounder on average than those of pure familial dextrals. To explain the difference, we propose that the cognitive effort of using exact numbers is greater for the familial sinistral group because their language and number systems tend to be more distributed over both hemispheres of the brain. Our data support the view that exact and approximate quantities are processed by two separate cognitive systems. Specifically, our behavioral data corroborates the view that the evolutionarily older, approximate number system is present in both hemispheres of the brain, while the exact number system tends to be localized in only one hemisphere.