The distribution of aspiration in Basque — with ‘aspiration’ referring to both the occurrence of [h] and of aspirated stops — shows some puzzling aspects. In some words, aspiration is ancient, in the sense that it must be assumed for the earliest reconstructable stage. In some other instances, however, it has arisen seemingly ex nihilo, as can be observed in borrowings from Latin and Romance, e.g. Latin/Romance īra &gt; Basque hira ‘ire’, Romance taula &gt; Basque thaula ‘board’. Most surprisingly, in some words aspiration has developed after a sonorant consonant, e.g. Romance solatz &gt; Basque solhas ‘conversation’. Aspiration may also continue intervocalic /n/, e.g. Latin anāte &gt; Basque ahate ‘duck’. Another unusual development is the phonologization of the contrast between aspirated and unaspirated voiceless stops triggered by a shift of the stress in some words without affecting the properties of consonants. Finally, an interdialectal alternation /k-/ ~ /g-/ ~ /h-/ ~ Ø in demonstratives and related adverbs appears to have involved fortition, contrary to initial expectations. Here we describe the environments in which aspiration is found in Basque and discuss the most likely historical developments that could have given rise to the state of affairs that we find, paying particular attention to what would appear to be unusual or unnatural sound changes. We build on prior scholarship, but this paper also contains some new hypotheses, especially regarding the aspiration in words like ahate ‘duck’.  We have also tried to contribute to the dating of the different processes and to the understanding of in their causes. José Ignacio Hualde 12 In (5) demonstratives and related forms, as well as borrowings, have been separated from other words with initial aspiration. The reason for this is that in both of these groups of words we can make use of additional evidence regarding the origin of the aspiration. Starting with the loanwords, it should be kept in mind that both Old Spanish and Gascon have phonemic aspiration, mostly because of a sound change whereby Latin /f/ became /h/ (aspiration was later lost in most Spanish dialects). Thus, in some of the borrowings in (5a), including hornitu ‘fournish’, from Gascon hornir (cf Fr fournir), hondar ‘bottom, dregs’, possibly from Spanish hondo ‘bottom’ (< Lat fundu-), and horma ‘ice’, from either Gascon or Spanish (< Lat forma), the aspiration can be attributed to the source Romance language. In most of the words in (5a), however, the aspiration lacks etymological justification, since these words are vowel-initial in Romance (see Michelena [1977] 2011, VI, 172). Although in harea from arēna the initial aspiration could perhaps be related to the weakening of intervocalic /n/ (see below), as proposed by Lakarra (2009a, 2009b, 2017), this explanation would not extend to other items in this list like harma ‘weapon’, hira ‘ire’, harrapatu ‘grab’ (< Gasc arrapar), husu ‘often, usually’, etc. We must thus conclude that some words have acquired non-etymological aspiration, as we concluded in the previous section regarding Romance borrowings with postconsonantal aspiration. The specific context for this prothesis of /h/ is, nevertheless, unclear, since it lacks systematicity. It appears to have been a sporadic process.10 In the native vocabulary, of course, we do not have the same evidence as in loanwords to determine which words may have developed non-etymological aspiration. In any case, from the examination of the distribution of initial /h/ in loanwords, we may safely assume that at least some of the native words in (5c) may have a non-etymological /h/, which does not continue any other older consonant. Trask (1997, 158–159) holds the view that aspiration in Basque, with very few potential exceptions, was originally a suprasegmental feature possibly related to accentual prominence. Gorrochategui & Lakarra (2001), on the other hand, have argued that this position is 10 It has been suggested that in some cases an initial /h/ may be due to contamination, e.g. harroka ‘rock’ (< Gasc arròca) may owe its aspiration to harri ‘stone’ (Gavel 1920, 190, Michelena [1977] 2011, VI, 172, but cf. also Gasc garròc). Lakarra (2015) discusses several such possible cases. 21 Aspiration in Basque allophonically aspirated in accented syllables and unaspirated in other syllables. The same rule affected the distribution of /h/: it was preserved only in the onset of an accented syllable, and words that did not have any aspiration tended to acquire it before the accented vowel in the contexts where it was licensed, as we can observe in loanwords, e.g. Lat īra > Bq */íra/ > hira ‘ire’; Rom solatz > Bq */solás/ > solhas ‘conversation’; Rom (h)onest > Bq */onést/ > onhest ‘honest’; Lat (h)onōrem > Bq */onóre/ > */oŋóre/ > */oŋhóre/ > ohore ‘honor’; Rom taula > Bq */táwla/ > thaula.17 As Michelena argues, a subsequent reinterpretation of the directionality of the accentual rule would have caused the phonologization of aspirated stops, by making their distribution not totally predictable from the position of the accent. Paradigm uniformity seems to have helped to preserve aspiration in syllables that were no longer stressed after this change, as in e.g. ákher ‘he-goat’, cf. inflected akhérra ‘the he-goat’, leading to phonologization of this feature. 9 Appendix: on the relationship between Aquitanian and Basque The distribution of the grapheme <h> in the Aquitanian inscriptions has been considered an essential piece of information regarding the history of aspiration in Basque. It is thus important to clarify the relationship between Basque and Aquitanian. The term ‘Aquitanian’ is commonly employed to refer to the language of non-Indo-European anthroponyms and theonyms that have been found in Latin inscriptions mostly in an area north of the Pyrenees to the East of the present-day Basque-speaking region, where the local language is Gascon. This area falls within the Roman