Gunnar Hansson

This dissertation provides a novel perspective on neutrality in vowel harmony, using evidence from multiple front/back and ATR harmony systems. While many standard accounts of harmony assume an equivalence between vowels that are neutral to harmony and those that lack a counterpart in the harmonic feature (e.g. van der Hulst 2016), this correspondence is demonstrably false in both directions. For example, in Hungarian (Chapter 3), [e:] lacks a harmonic counterpart, but is not consistently neutral to front/back harmony, in that it can alternate harmonically in some suffixes with [a:]. Conversely, in Mayak (Chapter 9), [a] has a contrastive ATR counterpart, yet is nonetheless neutral to ATR harmony. I argue that these types of patterns force a new, target-focused approach, where participation is based on the drive of specific vowel qualities to undergo harmony; neutrality results when this drive is insufficient to force unfaithfulness. This idea is motivated by cross-linguistic and phonetic facts suggesting that vowels that are low and/or rounded are inherently better targets of front/back harmony, while higher vowels are better targets of ATR-dominant harmony. I implement this approach formally in Harmonic Grammar; the harmony constraint is scaled by the quality of a vowel as a potential target, parallel to Kimper’s (2011) trigger strength scaling. This account can capture the complexities in the relationship between contrast and neutrality in a variety of harmony systems, including the gradience of neutrality (the height effect) in Hungarian (Hayes & Londe 2006), and paired neutral vowels in Mayak (Andersen 1999), among other cases. I argue that this view of harmony is necessary: neutrality is crucially about the quality of a vowel as a potential target of harmony, where target quality is determined in a cross-linguistically consistent, phonetically motivated way.

View record

A major question in phonology concerns the role of historical changes in shaping the typology of languages. This dissertation explores the effect of sound change on consonant inventories. Historical reconstruction is mainly done by comparing cognate words across languages, making it difficult to track how inventories change specifically. Additionally, few languages have historical written records that can be directly examined. For this dissertation, the main research tool is computer simulation, using bespoke software called PyILM, which is based on the Iterated Learning Model (Kirby 2011, Smith et al. 2003). This allows for the simulation of sound change from arbitrary starting points, controlling for a multitude of variables. PyILM is an agent-based model, where a 'speaking' agent transmits a set of words to a 'listening' agent. The speaking agent is then removed, the learner becomes the speaker, and a new learner is introduced. The cycle repeats any number of times, roughly simulating the transmission of language over many generations. Sound change in a simulation is due to channel bias (Moreton 2008), the result of which is that agents occasionally misinterpret some aspect of speech, and internalize sound categories that differ from the previous generation (Ohala 1981, Blevins 2004). Three typological generalizations are examined, none of which have previously been studied from an evolutionary perspective:(1) The total number of consonants in a language. This is shown to be related to syllable structure, such that languages with simple syllables develop smaller inventories than languages with complex syllables. This mirrors a positive correlation between inventory size and syllable structure in natural languages, as reported by Maddieson (2007).(2) The correlation reported by Lindblom and Maddieson (1988) between the size of an inventory and the complexity of its segments. This effect emerges in simulations when context-free changes are introduced, since these changes produce similar outcomes in inventories of all sizes.(3) Feature economy (Clements 2003), which refers to the way that consonants within a language tend to make use of a minimal number of distinctive features. Economy emerges over time when sound changes take scope over classes of sounds, rather than targeting individual sounds.

View record

Sound systems are a basic building block of any human language. An integral part of the acquisition of sound systems is the learning of allophony. In sound systems, some segments are used as allophones, or contextually-conditioned variants of a single phoneme, and learners need to figure out whether given segments are different phonemes or allophones. There is a growing interest in the question of how allophony is learned from speech input (e.g., Seidl and Cristia, 2012). This dissertation investigates the mechanisms behind the learning of allophony. Whether given segments are different phonemes or allophones of a single phoneme is partly determined by the contextual distribution of the segments. When segments occur in overlapping contexts and their occurrences are not predictable from the contexts, they are likely to be different phonemes. When segments occur in mutually exclusive contexts, and their occurrences are predictable from the contexts (i.e., they are in complementary distribution), the segments are likely to be allophones. This dissertation starts with the hypothesis that allophony can be learned from the complementary distribution of segments in input. With data collected in a series of laboratory experiments with adult English speakers, I make the following claims. First, adults can learn allophony from the complementary distribution of segments in input. The results of Experiment 1 showed that participants learned to treat two segments as something like allophones when they were exposed to input in which the segments were in complementary distribution. Second, the learning of allophony is constrained by the phonetic naturalness of the patterns of complementary distribution. The results of Experiment 2 showed that the learning of allophony happened only when participants were exposed to input in which relevant segments occurred in phonetically natural complementary contexts. Third, the learning of allophony involves the learning of the context-dependent perception of relevant segments. The results of Experiment 3 showed that, through exposure to input, participants’ perception of the relevant segments became more dependent on context such that they perceived the segments as being more similar to each other when they heard the segments in phonetically natural complementary contexts.

View record

An important property of any language’s sound system is its phonotactics—the unique way in which it allows its inventory of speech sounds to combine. Interestingly, certain types of phonotactic co-occurrence restrictions found in natural languages may hold across any amount of intervening material. For example, the Samala (Chumash) language of Southern California exhibits a pattern of sibilant harmony, such that [s] and [ʃ] may not co-occur anywhere within the same word (e.g. /ha-s-xintila-waʃ/ becomes [ha-ʃ-xintila-waʃ] `his former gentile name'; Applegate, 1972). Long-distance dependencies like this, despite being relatively common cross-linguistically, are known to pose serious problems for learnability. A learner needs an enormous amount of computational power to discover an interaction in an unbounded search space defined by arbitrary distances, resulting in patterns that are not learnable in practice. Their existence in natural languages thus suggests that humans are equipped with cognitive learning biases that restrict the available hypothesis space and facilitate the learning of patterns with certain properties but not others. This dissertation presents a series of artificial language learning studies that support the hypothesis that the typology of locality relations in long-distance consonantal phonotactics is shaped, at least in part, by such biases. From a theoretical perspective, the goal is to explore and define the boundaries of the human learner's hypothesis space for phonotactic patterns. I argue that the seemingly simple constraints used in the Agreement by Correspondence framework (Rose and Walker, 2004; Hansson, 2010; Bennett 2013) generate many pathological patterns that are unattested cross-linguistically. By contrast, the properties of locality observed for patterns of long-distance consonant agreement and disagreement belong to a well-defined and relatively simple class of subregular formal languages (stringsets) called the Tier-based Strictly 2-Local languages (TSL₂; Heinz et al., 2011). I therefore argue that class of TSL₂ stringsets offers an excellent approximation of the boundaries of possible, human-learnable phonotactics. More generally, I suggest that the formal-language-theoretic approach can be used to inform phonological theory, allowing for a better understanding of the computational complexity and learnability of predicted patterns.

View record

This thesis investigates how Shona, an African language spoken in Zimbabwe deals with potentially onsetless syllables (heterosyllabic VV sequences & initial onsetless syllables) and subminimal words. The thesis focuses on the morphophonemics of Karanga and Zezuru—the two principal dialects of Shona. Karanga and Zezuru morphophonemic processes observed in this thesis have only one primary goal; to achieve the typical or preferred Shona phonological structures—the consonant-vowel (CV) syllable and the disyllabic Prosodic Word. Often, when morphemes are concatenated, the resultant phonological structures do not conform to these typical structures. The study examines the repair strategies that Karanga and Zezuru employ to achieve the CV syllable and the disyllabic Prosodic Word. The overall analysis is couched in Optimality Theory (Prince and Smolensky (2004 [1993].Hiatus resolution strategies are conditioned by prosodic domains/boundaries, and a detailed prosodic parsing is required to account for this phenomenon. The Prosodic Stem, Prosodic Word and the Clitic Group are the prosodic domains relevant for this study. Owing to the impossibility of unifying the cliticization and coalescence facts with the other strategies in a single constraint ranking, two strata are posited—the Word (lexical) and the Phrasal (Postlexical) using the Lexical Phonology and Morphology-Optimality Theory (Kiparsky 2000, 2003). At the Word level, Glide formation is the default strategy, and at the Phrasal level, it is coalescence. Employing the Clements and Hume (1995) Unified Feature Geometry model, with the addition of the feature [pharyngeal], all the hiatus-breakers [j w ʔ ɦ] are analyzed as products of spreading.Karanga and Zezuru display greater variation with respect to prosodic minimality and initial onsetless syllables than in hiatus resolution. It is argued that Zezuru enforces WORD MINIMALITY at the expense of ONSET, and Karanga enforces ONSET at the expense of WORD MINIMALITY. Karanga displays internal variation; it allows initial onsetless syllables in function words but not in lexical ones. Based on tone, reduplication, minimality and cliticization, initial onsetless syllables are argued to be morified, syllabified and not extra-prosodic and therefore do not warrant any special representation.

View record

This thesis deals with the phonology of Gitksan, a Tsimshianic language spoken innorthern British Columbia, Canada. The claim of this thesis is that Gitksan exhibits severalgradient phonological restrictions on consonantal cooccurrence that hold over the lexicon.There is a gradient restriction on homorganic consonants, and within homorganic pairs, thereis a gradient restriction on major class and manner features. It is claimed that theserestrictions are due to a generalized OCP effect in the grammar, and that this effect can berelativized to subsidiary features, such as place, manner, etc. It is argued that these types ofeffects are best analyzed with the system of weighted constraints employed in HarmonicGrammar (Legendre et al. 1990, Smolensky & Legendre 2006).It is also claimed that Gitksan exhibits a gradient assimilatory effect among specificconsonants. This type of effect is rare, and is unexpected given the general conditions ofdissimilation. One such effect is the frequency of both pulmonic pairs of consonants andejective pairs of consonants, which occur at rates higher than expected by chance. Another isthe occurrence of uvular-uvular and velar-velar pairs of consonants, which also occur at rateshigher than chance. This pattern is somewhat surprising, as there is a gradient prohibition oncooccurring pairs of dorsal consonants. These assimilatory patterns are analyzed using theAgreement by Correspondence approach (Hansson 2001, Rose & Walker 2004), whichmandates that output correspondents agree for some phonological feature.The general discussion of assimilation and dissimilation is continued inmorphological contexts, such as reduplication. It is claimed there are differences in thegradient and categorical patterns of assimilation and dissimilation in Coast Tsimshian andGitksan reduplicative contexts. A summary of the attested reduplicative patterns in thelanguages, as well as results from a nonce-probe task, supports this claim. This differencebetween Coast Tsimshian and Gitksan is indicative of a larger difference in the reduplicativepatterns of the languages of the Tsimshianic family: each member of the family exhibitsslightly different patterns of deglottalization. A typological study of these patterns suggeststhat glottalized sonorants and obstruents are fundamentally different segment types.

View record