I present evidence from Navajo and English that weaker, gradient versions of morpheme-internal phonotactic constraints, such as the ban on geminate consonants in English, hold even across prosodic word boundaries. I argue that these lexical biases are the result of a maximum entropy phonotactic learning algorithm that maximizes the probability of the learning data, but that also contains a smoothing term that penalizes complex grammars. When this learner attempts to construct a grammar in which some constraints are blind to morphological structure, it underpredicts the frequency of compounds that violate a morpheme-internal phonotactic. I further show how, over time, this learning bias could plausibly lead to the lexical biases seen in Navajo and English.

In languages with strident harmony, stridents within a particular domain are required to have the same minor place of articulation. Harmony is often required only of stridents within a root or stem morpheme, and doesn't trigger alternations. Harmony is also often quite local, applying exclusively or more strongly between stridents in the same or adjacent syllables. Finally, harmony may be morpheme specific, triggering alternations in some affixes but not others. All of these specifics of a given harmony pattern give rise to exceptions to harmony at the level of the word, and may require a morphologically parsed learning corpus in order to be acquired. This paper explores the learnability of strident harmony in text corpora from three languages: Nkore-Kiga (Bantu), Papantla Totonac (Totonacan) and Navajo (Athapaskan). The analyses show that word level exceptions largely obscure the harmony pattern as an overall phonotactic in a language. The three languages also serve as a test of the Projection Induction Learner (Gouskova & Gallagher 2020), which is found to be successful when the generalizations in the data are strong but may fail in the face of patterned exceptions.

Saltatory alternations occur when two sounds alternate with each other, excluding a third sound that is phonetically intermediate between the two alternating sounds (e.g. [p] alternates with [β], with nonalternating, phonetically intermediate [b]). Such alternations are attested in natural language, so they must be learnable; however, experimental work suggests that they are dispreferred by language learners. This article presents a computationally implemented phonological framework that can account for both the existence and the dispreferred status of saltatory alternations. The framework is implemented in a maximum entropy learning model (Goldwater & Johnson 2003) with two significant components. The first is a set of constraints penalizing correspondence between specific segments, formalized as *MAP constraints (Zuraw 2007, 2013), which enables the model to learn saltatory alternations at all. The second is a substantive bias based on the P-map (Steriade 2009 [2001]), implemented via the model's prior probability distribution, which favors alternations between perceptually similar sounds. Comparing the model's predictions to results from artificial language experiments, the substantively biased model outperforms control models that do not have a substantive bias, providing support for the role of substantive bias in phonological learning.

This paper presents maxent.ot, a package for doing phonological analysis using Maximum Entropy Optimality Theory written in the statistical programming language R. R has become the de facto standard for doing statistical analysis in linguistic research, and this package allows phonologists to create and disseminate MaxEnt OT analyses in R. A central goal of the package is to support reproducible research and to allow the crucial components of a MaxEnt analysis to be performed conveniently and with only a basic knowledge of R programming. The paper first presents a tutorial on MaxEnt constraint grammars and how to use maxent.ot to perform a simple analysis. We then turn to more advanced features of the package, including model comparison, regularization, and cross-validation.

This article examines a case of phonological opacity in Uyghur resulting from an interaction between backness harmony and a vowel reduction process that converts harmonic vowels into transparent vowels. A large-scale corpus study shows that although opaque harmony with the underlying form of a reduced vowel is the dominant pattern, cases of surface-apparent harmony also occur. The rate of surface-apparent harmony varies across roots and is correlated with a number of factors, including root frequency. These data pose problems for standard accounts of opacity, which do not predict such variation. I propose an analysis where variation emerges from conflict between a paradigm uniformity constraint mandating that the harmonising behaviour of a root remains consistent, and surface phonotactic constraints. This is implemented in a parallel model by scaling constraint violations according to certainty in a root’s harmonic class. This aligns with past work suggesting some opacity is driven by paradigm uniformity.

We present a simple and robustly incentive-compatible price list methodology to elicit quantiles of a subjective real-valued belief. These elicited quantiles can be employed to approximate a subject’s complete subjective distribution, and we establish that the distribution maximizing entropy while adhering to the elicited quantiles is piecewise linear. Using this approach, our methodology extends to estimating arbitrary unobserved attributes of the subjective distribution, such as mean and variance, which are otherwise challenging to elicit. We provide a proof-of-concept for our framework through an experiment involving the elicitation of participants’ beliefs regarding the mathematical abilities of their peers.

The dhole Cuon alpinus is a large canid that is categorized as Endangered on the IUCN Red List and at risk of global extinction. Information on the spatial distribution of suitable habitat is important for conservation planning but is largely unavailable. We quantified the spatial distribution of potential range as well as the relative probability of dhole occurrence across large parts of the species’ global range. We used the MaxEnt algorithm to produce a multi-scale environmental niche model based on 24 environmental variables and dhole occurrence data from 12 countries. We identified three regions where dhole conservation should be focused: western India, central India, and across the Himalayan foothills through Southeast Asia. Connectivity between suitable areas was poor, so coordinated action among these regions should be a priority. For instance, transboundary dhole conservation initiatives across the Himalayas from southern China, Myanmar, north-east India, Nepal and Bhutan need to be initiated. We also highlight the value of improving dhole population viability on unprotected land and increasing monitoring in the northern parts of its historic distribution, in particular in areas within mainland China.

The rare Himalayan wolf Canis lupus chanco is categorized as Vulnerable on the IUCN Red List, and there is limited knowledge of its ecology and distribution. In Bhutan, the Himalayan wolf is one of the least known carnivores. Our aims in this study were to map the current distribution of the wolf in Bhutan and to identify potential habitats within the country. We compiled 32 records of wolf presence from camera-trap surveys and, using a maximum entropy approach, we estimated a potential habitat of 2,431 km2, comprising c. 6.3% of Bhutan. However, wolf presence was localized and non-continuous. We recommend a detailed fine-scale habitat analysis in areas of potential habitat and genetic studies to investigate population structure. Knowledge of these matters will provide insights regarding connectivity and facilitate the development of conservation strategies for this threatened wolf.

Several authors have investigated the question of whether canonical logic-based accounts of belief revision, and especially the theory of AGM revision operators, are compatible with the dynamics of Bayesian conditioning. Here we show that Leitgeb’s stability rule for acceptance, which has been offered as a possible solution to the Lottery paradox, allows to bridge AGM revision and Bayesian update: using the stability rule, we prove that AGM revision operators emerge from Bayesian conditioning by an application of the principle of maximum entropy. In situations of information loss, or whenever the agent relies on a qualitative description of her information state—such as a plausibility ranking over hypotheses, or a belief set—the dynamics of AGM belief revision are compatible with Bayesian conditioning; indeed, through the maximum entropy principle, conditioning naturally generates AGM revision operators. This mitigates an impossibility theorem of Lin and Kelly for tracking Bayesian conditioning with AGM revision, and suggests an approach to the compatibility problem that highlights the information loss incurred by acceptance rules in passing from probabilistic to qualitative representations of belief.

Habitat prioritization and corridor restoration are important steps for reconnecting fragmented habitats and species populations, and spatial modelling approaches are useful in identifying suitable habitat for elusive tropical rainforest mammals. The Endangered Bornean banteng Bos javanicus lowi, a wild bovid endemic to Borneo, occurs in habitat that is highly fragmented as a result of extensive agricultural expansion. Based on the species’ historical distribution in Sabah (Malaysia), we conducted camera-trap surveys in 14 forest reserves during 2011–2016. To assess suitable habitat for the banteng we used a presence-only maximum entropy (MaxEnt) approach with 11 spatial predictors, including climate, infrastructure, land cover and land use, and topography variables. We performed a least-cost path analysis using Linkage Mapper, to understand the resistance to movement through the landscape. The surveys comprised a total of 44,251 nights of camera trapping. We recorded banteng presence in 11 forest reserves. Key spatial predictors deemed to be important in predicting suitable habitat included soil associations (52.6%), distance to intact and logged forests (11.8%), precipitation in the driest quarter (10.8%), distance to agro-forest and regenerating forest (5.7%), and distance to oil palm plantations (5.1%). Circa 11% of Sabah had suitable habitat (7,719 km2), of which 12.2% was in protected forests, 60.4% was in production forests and 27.4% was in other areas. The least-cost path model predicted 21 linkages and a relatively high movement resistance between core habitats. Our models provide information about key habitat and movement resistance for bantengs through the landscape, which is crucial for constructive conservation strategies and land-use planning.

In modelling joint probability distributions it is often desirable to incorporate standard marginal distributions and match a set of key observed mixed moments. At the same time it may also be prudent to avoid additional unwarranted assumptions. The problem is to find the least ordered distribution that respects the prescribed constraints. In this paper we will construct a suitable joint probability distribution by finding the checkerboard copula of maximum entropy that allows us to incorporate the appropriate marginal distributions and match the nominated set of observed moments.

Throughout the world, the invasion of nonnative plants is an increasing threat to native biodiversity and ecosystem sustainability. Invasion is especially prevalent in areas affected by land transformation and disturbance. Surface mines are a major land transformation, and thus may promote the establishment and persistence of invasive plant communities. Using the Shale Hills region of Alabama as a case study, we assessed the use of landscape characteristics in predicting the probability of occurrence of six invasive plant species: sericea lespedeza, Japanese honeysuckle, Chinese privet, autumn-olive, royal paulownia, and sawtooth oak. Models were generated for invasive species occurrence using logistic regression and maximum entropy methods. The predicted probabilities of species occurrence were applied to the mined landscape to assess the probable prevalence of each species across the landscape. Japanese honeysuckle had the highest probable prevalence on the landscape (48% of the area), with royal paulownia having the lowest (less than 1%). Overall, 67% of the landscape was predicted to have at least one invasive plant species, with 20% of the landscape predicted to have two or more species, and 3% of the landscape predicted to have three or more species. Japanese honeysuckle, sericea lespedeza, privet, and autumn-olive showed higher occurrence on the reclaimed sites than across the broader region. We found that geospatial modeling of these invasive plants at this scale offered potential for management, both for identifying habitat types at risk and areas that need management attention. However, the most immediate action for reducing the prevalence of invasive plants on reclaimed mines is to remove invasive plants from the reclamation planting list. Three (sericea lespedeza, autumn-olive, and sawtooth oak) out of the six most common invasive plants in this study were planted as part of reclamation activities.

Data on the number of Pennsylvania dairy farms by size category are analyzed in a Markov chain setting to determine factors affecting entry, exit, expansion, and contraction within the sector. Milk prices, milk price volatility, land prices, policy, and cow productivity all impact structural change in Pennsylvania's dairy sector. Stochastic simulation analysis suggests that the number of dairy farms in Pennsylvania will likely fall by only 2.0 percent to 2.5 percent annually over the next 20 years, indicating that dairy farming in Pennsylvania is likely to be a significant enterprise for the state in the foreseeable future.

Juggler's exclusion process describes a system of particles on the positive integers where particles drift down to zero at unit speed. After a particle hits zero, it jumps into a randomly chosen unoccupied site. We model the system as a set-valued Markov process and show that the process is ergodic if the family of jump height distributions is uniformly integrable. In a special case where the particles jump according to a set-avoiding memoryless distribution, the process reaches its equilibrium in finite nonrandom time, and the equilibrium distribution can be represented as a Gibbs measure conforming to a linear gravitational potential.

Let S: [0, 1]→[0, 1] be a chaotic map and let f* be a stationary density of the Frobenius-Perron operator PS : L 1→L 1 associated with S. We develop a numerical algorithm for approximating f*, using the maximum entropy approach to an under-determined moment problem and the Chebyshev polynomials for the stability consideration. Numerical experiments show considerable improvements to both the original maximum entropy method and the discrete maximum entropy method.

The distribution of the Near Threatened Caucasian grouse Tetrao mlokosiewiczi, endemic to the Caucasus, was examined to model the species’ nesting habitat, and thus facilitate its conservation and the identification of Key Biodiversity Areas in the Caucasus. The species’ occurrence was defined by field surveys and radio-telemetry. Data were managed and analysed using a geographical information system and various modelling techniques. Grouse locations were divided into training and testing datasets. Habitat variables measured at training locations were used to develop models, and testing locations were used to validate the models. The final best-fit model suggested that Caucasian grouse prefer open habitat, and the most important independent variables accounting for the species' distribution were annual mean temperature, mean temperature of warmest quarter, precipitation seasonality and proximity to deciduous broad-leaf forest. The incorporation of human disturbance and ruggedness into the final model significantly increased its predictive power. This model provides a tool to improve search effectiveness for Caucasian grouse in the Caucasus and for the conservation and management of the species. The model can predict the probable distribution of Caucasian grouse and the corridors between known populations. Threatened and endemic species are often used as species for setting site-based conservation priorities, and this habitat model could help to identify new Key Biodiversity Areas for protection in the Caucasus. The Ministry of Environmental Protection and Natural Resources of Georgia is going to use the results of this study to reshape existing protected areas and identify new ones.