Division of Psychology and Language Sciences

This paper presents the results of a survey on the attitudes of young British people (age range 18-25) towards the learning of the UK’s regional languages. A combined quantitative and qualitative questionnaire was administered to 71 respondents, who were all born and spent the majority of their youth in Britain. It was found that overall attitudes to regional languages are largely positive and that a great proportion of the target group reported they were potentially willing to learn one of the regional languages. Cultural identity and a desire to integrate with the community were identified as major contributors to willingness to learn. A major factor for unwillingness to learn was “lack of real need” (i.e. the ability to get by with English). The paper also looks at the distribution of these attitudes against gender, age and education, all of which were significantly related. Further, respondents already speaking a regional language were found to have more favourable attitudes overall.

This study looks at how the borrowed swearwords ffwcin ‘fucking’ and blydi ‘bloody’, which like prif ‘main’ and hen ‘old’ occur in the non-default modifier-head position in colloquial Welsh, are integrated into the initial consonant mutation environment traditionally associated with this. It is proposed that speakers can follow three different strategies in integrating these new lexical items with their existing system: use the same patterns as for existing items in the category, treat them as invisible to the mutation process, or do not apply mutation at all. Hypotheses were then tested in a translation based informant session with seven participants and it was concluded that participants do indeed use all three strategies in integrating these items into their existing systems, but do so with high idiosyncracy and even show intra-speaker variation.

"A frequent argument for an element-theoretic model of segmental representation (KLV 1985, Harris 1994, Harris & Lindsey 1995) is that its smaller generative capacity provides a better fit to the attested data across languages, while classical feature-based models suffer from significant overgeneration (e.g. Chen 2010, Backley 2011). However, little attempt has been made to provide a direct quantitative comparison of the generative capacities of the two approaches.

Reiss (2012) argues that overgeneration is actually a desirable property for a theory of segmental representation. His argument is based on the wide range of observed variation in size and content of languages’ sound inventories, and the assumption that no generalization about the impossibility of further possible phones can be made simply from the fact that a phone is as of yet unattested.

In this paper I take a contrasting stance and argue for distinguishing different factors in evaluating the generative capacity of segmental theories: possible segments, possible inventories and possible natural classes. Only the prediction of a large number of inventories is desirable and necessary, while the sets of possible segments and natural classes should be accounted for restrictively. This reflects the fact that the sizes of individual inventories are limited (Crystal 2010 and the UPSID give between 11 phonemes for Rotokas and Pirahã to 141 in !Xũ) and that a model that needs to appeal to fewer natural classes to account sufficiently for phonological processes is more elegant and parsimonious.

Building on the work of Reiss (2012), I formally compare the generative capacities of Element Theory (ET), Feature Theory (FT) and Underspecification Theory (UT). I show that FT is less powerful than ET and ET less powerful than UT on all three measures, but that considering the number and nature of the primitives (viz. features and elements) each theory relies upon, ET provides the best fit to the above criteria.

References:
Backley, P. 2011. An Introduction to Element Theory. Edinburgh: Edinburgh University Press
Chen, Y.-L. 2010. How Powerful Are Elements? An Evaluation of the Adequacy of Element Theory in Phonological Representations. MSc Dissertation. University of Edinburgh (Available: http://hdl.handle.net/1842/5333)
Crystal, D. 2010. The Cambridge Encyclopedia of Language (3rd Edition). Cambridge: Cambridge University Press, p. 173
Harris, J. 1994. English Sound Structure. Oxford: Blackwell
Harris, J. & Lindsey, G. 1995. The elements of phonological representation. In: Durand, J. & Katamba, F. (Eds.), Frontiers of phonology: atoms, structures, derivations. Harlow, Essex: Longman, pp. 34-79
Kaye, J., Lowenstamm, J. and Vergnaud, J.-R. 1985. The internal structure of phonological elements: a theory of charm and government. Phonology, 2. pp. 305-328
Reiss, C. 2012. Towards a bottom-up approach to phonological typology. In: Di Sciullo, A. M. (Ed.) Towards a Biolinguistic Understanding of Grammar: Essays on Interfaces. Amsterdam: John Benjamins"

As one of the main contenders of the theory of distinctive features, Element Theory (ET) has seen much change over the last decades. While this has vastly advanced its capability and accuracy as a theory of subsegmental phonology, on-going development has sometimes come at the price of explicitness in definition of the precise model assumed and consequently there are some unclarities and apparent contradictions in some of the current proposals of ET.
This dissertation first gives an outline of the current state of ET in light its historical development, highlighting the ways in which it differs from Feature Theory, Autosegmental Phonology and Government Phonology. The position that it functions as a relatively independent model of subsegmental representation is advanced.
On the basis of this, the dissertation proposes a concrete formalisation of ET, ground in basic mathematical set theory. It is argued that, somewhat analogous to the set-representation of syntactic treelets as ordered pairs {α, {α, β }}, segmental representations can be seen as partially ordered sets of the type {{α}, {α, β, γ, ...}}. It is illustrated how such sets can be used as the basis to formalise the model of ET set out here, including aspects such as composition, decomposition, well-formedness, and geometry.
Finally, based on the work of Reiss (2012), this model is used to compare the generative capacity and power of ET to that of classical feature theory. It is argued that a distinction needs to be made between generative capacity and generative power, and that the desirability or undesirability of overgeneration/powerfulness is more fine-grained and differs between the individual aspects of subsegmental phonology, specifically the set of all possible segments, the set of all possible inventories and the set of all natural classes. It is shown that ET is, converse to common assumption, actually more powerful than feature theory but its concrete capacity is reduced by the relatively small number of primes assumed.

Most recent work in Element Theory assumes that nasality and true voicing are represented by the same element, where the headed element encodes voicing and the dependent element nasality (Backley, 2011; Nasukawa, 1999, 2000, 2005, et alii). This assumption is questioned here and it is proposed that the voicing–nasality contrast may be encoded the other way around. It is argued that this hypothesis is in better agreement with the means by which headedness may encode additional information at phonetic interpretation. Under the assumption of an L/H-Parameter, it is shown how, in line with Lombardi (1991) and Botma (2005), only H-systems can possibly encode voiceless (or more properly aspirated) nasals and how the hypothesis that headed |L| encodes nasality provides a better fit for these systems.

"Background. While it is well known from colloquial Welsh that speakers frequently omit the initial auxiliary in periphrastic constructions of the form AuxSVO, this phenomenon has not received much explicit attention in the literature. An example of a sentence with auxiliary deletion (AD) is given in (1) below, with the deleted auxiliary in parenthesis.
(1) (Wyt) ti ddim yn licio dreifio?
    (be.2S) 2S NEG IMPFV like drive
    ‘You don’t like driving?’ (Davies 2010:267, my gloss)
Jones (2004) primarily discusses how the optionality of the initial copula in periphrastic constructions with bod ‘to be’ is limited by the grammatical person of the clause’s subject. Davies (2010) and Davies & Deuchar (in prep.) present the first empirical evaluation of AD by means of studying data from the Siarad 1 corpus. They offer as a possible explanatory hypothesis convergence toward an English-like SVO surface structure which is achieved by omission of the auxiliary, but limit their study to AD with the second person familiar singular ti. The phenomenon is also briefly mentioned in Borsley et al. (2007), but they do not offer a more in-depth analysis other than observing which grammatical persons AD may occur with, similar to Jones (2004) but more limited and with slightly different accounts of acceptability. It is not currently known whether factors other than grammatical person constrain the occurrence of AD and there is no consensus in the literature on the range of persona AD may occur with. It is even unclear whether AD is limited more strictly to deletion of the copula bod ‘to be’ or is indeed applicable to the wider range of auxiliaries including ddaru ‘PAST’ and gwneud ‘to do’. Employing a corpus study and an acceptability experiment, this study thus seeks to establish (a) in how far AD is constrained by grammatical person, and (b) whether a number of other factors such as type of auxiliary, tense, aspect and coordination constrain AD beyond this.

Method. To establish the range of use of AD with different grammatical persons, 1662 instances of AD were programmatically extracted from the Siarad corpus. For this the entire corpus was first processed with the Bangor AutoGlosser 2 to obtain a tier of uniform glosses. Because clause initial verbal structure could not be relied upon to identify sentences with AD, a turn-based approach was taken. A script then extracted data about every turn in which the first token encountered was a pronoun with person and number marking. To test the factors given under (b) above, a bi-modal judgement experiment with a battery of stimuli testing AD under varying conditions was employed. For every condition tested (other than grammatical person and number), four pairs of stimuli were presented, one with AD and one without, as for instance in (2) below to test for bod plus past tense:
(2) (AD) Ti’n siopa am oriawr dwy flynedd yn ôl.
    (Test) Roeddet ti’n siopa am oriawr dwy flynedd yn ôl.
    ‘You were shopping for a watch two years ago.’
The entire set of stimuli was recorded by a female native speaker of Welsh. 20 participants (ages 19 to 58, M=31.85) were then first presented with an auditory judgement task. This involved participants being presented with two instances of the randomised entire set of recorded stimuli over headphones and forced to decide whether what they heard sounded acceptable to them or not in a limited amount of time, after which they were automatically presented with the next stimulus. Following this task, participants were presented with a written version of the randomised entire set of stimuli accompanied by a Likert scale (range 1 to 5) to indicate perceived acceptability.

Results. Results from the corpus study show that while the predictions made by both Jones (2004) and Borsley et al. (2007) generally fit quite well, especially in that they both allow AD with the second person singular and not with the third person singular, AD with first person singular is more common than expected and all other categories appear to be more limited. The subsequent judgement experiment appears to largely confirm the findings of the corpus study, but third person singular exhibited a significant gender bias, with masculine pronouns somewhat acceptable, though still notably below any comparable threshold to the others, and NPs as subjects yield very mixed results. The results further show that AD is only acceptable with the copula bod and not with gwneud, and that acceptability is sensitive to tense (permissible with present and future tense, but not past), while aspect appears to be insignificant. Likewise, the results exclude mood as a significant factor. While the results show that word order (AuxSVO v. SAuxVO v. VOAuxS v. OAuxSV) only has a small punitive effect on AD acceptability, test modality appears to play a role, penalising AD in written contexts with fronted subject (SAuxVO). Lastly, neither subordination nor Wh-movement (with and without pied-piped preposition) seem to negatively impact on AD acceptability.

Summary. This study further refines Jones’ (2004) and Borsley et al.’s (2007) observations of AD constraints linked to the grammatical person and number of the clause’s subject and suggests that tense is an additional factor not previously taken into account. It is clearly shown that AD is a case of copula omission, and not applicable to other auxiliaries. Finally, it is also shown that the results do not conflict with Davies’ (2010) hypothesis that this is change in progress and further avenues of research are suggested, especially toward understanding the underlying importance of constraining factors."

Initial Consonant Mutations (ICM), in which the initial segment of a word is phonologically altered in certain morphosyntactic contexts, are often thought of as a peculiarity of the Celtic languages. Moreover, the phonological implementation of these processes is a well-known challenge for phonological theory. This paper presents a refined analysis of one specific type of ICM, Aspirate Mutation (AM), in Welsh and Iaai. In Welsh the most basic form changes the voiceless plosives /p, t, k/ into corresponding fricatives, as exemplified in (1).
(1) ci [kiː] ‘dog’ → (â) chi [χiː] ‘with dog’
tŷ [tiː] ‘house’ → (ei) thŷ [θiː] ‘her house’
Note that the triggers (â, ei) can also be realised as enclitics to a preceding word, e.g. in o’i thŷ ‘from her house’. This means the effect is not contained by a clitic group (cf. Pyatt 2003). Previous element-theoretic work on AM by Buczek (1995) and Cyran (2010) has achieved much simplification over previous feature-theoretic work by reducing ICM patterns to the (de)composition of single elements. It has however also been documented that some speakers of Welsh extend AM to apply not only to the voiceless plosives, but also to the nasals /m, n/, which are altered to their voiceless counterparts [m̥, n̥] in this context (King 2003), as in (2).
(2) nain [naɪn] ‘grandmother’ → (ei) nain [n̥aɪn] ‘her grandmother’
ni [niː] ‘us’ → (â) ni [n̥iː] ‘with us’
While it is widely agreed that voiceless nasals (and nasal mutation) involve dependent |H| (Buczek 1995, Botma 2004, Cyran 2010, Backley 2011, Breit 2013b, et alii), the lenition-based accounts in Buczek (1995) and Cyran (2010) cannot account for the extension of AM to nasals. This relates principally to the common assumption that headed |L| represents voicing and dependent |L| nasality (Nasukawa 1999, 2000, 2005; Botma 2004). This paper shows how the opposite assumption, i.e. headed |L| for nasality and dependent |L| for voicing, enables us to take a fortition-based approach (note the strong word-initial environment), by composing headed |H| (representing frication) in the AM environment. Since the Single Optional Headedness Condition (cf. e.g. Breit 2013a:5) rules out the doubly-headed representation
*|L, H| for the targeted nasals, |H| can only be composed as a dependent, leading to realisation of a voiceless nasal. Loss of edge (|ʔ|) in plosives is due to an inability to license |ʔ| in the context of |H|. This is illustrated in (3a) for plosive → fricative and (3b) for voiced → voiceless nasal (neglecting place and edge).
(3) (a)  x  →  x (b)  x  →  x
      |    |               |    |
    H    H               L    L
                                |
                                H
    [t]    [θ]     [n]  [n̥]
Moreover, the exact same fortition-based analysis can be applied to Iaai, an Oceanic language which shows alteration of the initial consonant from voiced to voiceless nasals in certain morphosyntactic contexts such as object incorporation, as shown in (4).
(4) [naŋ] ‘brandish’ → [n̥aŋ] ‘brandish (incorp. obj.)’
[an] ‘eat’ → [han] ‘eat (incorp. obj.)’ (Maddieson & Anderson 1995)
This refined analysis thus unifies the alteration between plosives and fricatives, and voiced and voiceless nasals under AM in both Welsh and unrelated Iaai, for which such an analysis has not previously been proposed. The new analysis is also advantageous in positing that AM is fortition (i.e. addition of elemental content), consistent with the strong word-initial environment, rather than a form of lenition (i.e. loss of elemental content).