The Nerja Cave is a key archaeological site in the Southern Iberian Peninsula. It was inhabited by humans from the Upper Palaeolithic until recent Prehistory (30 and 3.7 ka cal BP). Various excavation campaigns performed in its external chambers (Vestíbulo, Mina and Torca) have recovered evidence of its use as habitat and burial site. Multiple studies on these matters have been published, but, until now, no Bayesian chronological modeling that utilized radiocarbon dates of the three chambers has been performed. To do so, all the available radiocarbon dates and stratigraphic and archaeological data have been compiled. These comprehend ample and diverse information about which, firstly, individual phase models based on the stratigraphic sequence of each one of the chambers have been created. After critically evaluating the results for each of the chambers, a general phase model for the prehistoric occupation of the external chambers has been created considering the cultural adscription of the samples. This has enabled the identification of 11 phases which correspond to the different technocomplexes of the Gravettian, Solutrean, Magdalenian, Epipalaeolithic, Mesolithic, Neolithic and Chalcolithic. Still pending are the refinement and improvement of the model for the Neolithic horizon among other phases of the sequence. The individual and the general models have evidenced important differences between the different archaeological phases in radiocarbon information as well as in the occupation of the three chambers.

We examine universals and crosslinguistic variation in constraints on event segmentation. Previous typological studies have focused on segmentation into syntactic (Pawley 1987) or intonational units (Givón 1991). We argue that the correlation between such units and semantic/conceptual event representations is language-specific. As an alternative, we introduce the MACRO-EVENT PROPERTY (MEP): a construction has the MEP if it packages event representations such that temporal operators necessarily have scope over all subevents. A case study on the segmentation of motion events into macro-event expressions in eighteen genetically and typologically diverse languages has produced evidence of two types of design principles that impact motion-event segmentation: language-specific lexicalization patterns and universal constraints on form-to-meaning mapping.

Despite the crucial dependence of synchronic meaning on both historical and cognitive context, we have traditionally used different tools for expressing synchronic and diachronic generalizations in modeling a complex semantic category like the diminutive. This is due in part to the extraordinary, often contradictory range of its senses synchronically (small size, affection, approximation, intensification, imitation, female gender), and the difficulty of proposing a coherent historical reconstruction for these senses.

I propose to model the synchronic and diachronic semantics of the diminutive category with a RADIAL CATEGORY (George Lakoff 1987), a type of structured polysemy that explicitly models the different senses of the diminutive and the metaphorical and inferential relations which bind them. Synchronically, this model explains the varied and contradictory senses of the diminutive. Diachronically, the radial category acts as a kind of ARCHAEOLOGY OF MEANING, expressing the generalizations of the classic mechanisms of semantic change (metaphor, abstraction and inference) as well as a new one: LAMBDA-ABSTRACTION, which accounts for the rise of quantificational meaning and second-order predicates in the diminutive. The model also predicts that the origins of the diminutive cross-linguistically lie in words semantically or pragmatically linked to children. I test the model by considering the semantics of the diminutive in over 60 languages, examining the origins of the diminutive in many of these, particularly in Indo-European where the theory suggests a new reconstruction of the proto-semantics of the PIE suffix -ko-.

Phonological constraints can, in principle, be classified according to whether they are natural (founded in principles of universal grammar (UG)) or unnatural (arbitrary, learned inductively from the language data). Recent work has used this distinction as the basis for arguments about the role of UG in learning. Some languages have phonological patterns that arguably reflect unnatural constraints. With experimental testing, one can assess whether such patterns are actually learned by native speakers. Becker, Ketrez, and Nevins (2007), testing speakers of Turkish, suggest that they do indeed go unlearned. They interpret this result with a strong UG position: humans are unable to learn data patterns not backed by UG principles.

This article pursues the same research line, locating similarly unnatural data patterns in the vowel harmony system of Hungarian, such as the tendency (among certain stem types) for a final bilabial stop to favor front harmony. Our own test leads to the opposite conclusion of Becker and colleagues': Hungarians evidently do learn the unnatural patterns.

To conclude we consider a bias account—that speakers are able to learn unnatural environments, but devalue them relative to natural ones. We outline a method for testing the strength of constraints as learned by speakers against the strength of the corresponding patterns in the lexicon, and show that it offers tentative support for the hypothesis that unnatural constraints are disfavored by language learners.

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.