Meeting 1 - Synopsis and Thoughts for the future
Heather Newell (UQAM)
N.B. This synopsis of the day is post-hoc, and liable to a view skewed by personal interests, to errors, and to omissions (some already corrected by Tobias). Please feel free to correct and/or add to this account in the comment section below. It would also be lovely if this synopsis enabled a continuation of the many points discussed at PTA1, by both people who attended and those who did not.
The first meeting of the CNRS-GDRI group, the Phonological Theory Agora (PTA), took place in Lublin, Poland, and was hosted by KUL (and they were spectacular hosts!).
The theme of the meeting was The relation between Representation and Derivation, and its aims were to discuss the theoretical implications of structural phonological representations and computational operations within the phonological component, or in other linguistic modules that have an impact on the phonology. Specific questions addressed were the following: What are the reasons for preferring representational over derivational solutions to a particular problem? How do representational and derivational theories relate to each other, how do they interact? For instance, can any kind of representation function in any theory of what a derivation is? To what extent are certain derivational solutions notational variants of representational ones, and vice versa?
The day started with a tutorial led by Tobias Scheer (Nice) and Marc van Oostendorp (Amsterdam, Leiden) on the distinction between representation and derivation (handout available online). This immediately began with a shift in terminology; derivation became computation in order to avoid the serial implications of the former. It then delved into what the properties of a computation were, leading to much debate and disagreement even between the tutorial leaders. Tobias proposed that a representation is something that could be ill-formed, which was not immediately accepted by the crowd and led to a short discussion of whether ill- formedness was actually possible. Another two definitions of a representation that came out over the day and a half were that a stored item in the lexicon must be a representation, and that a representation cannot have an input or an output. It was noted that storage is not specific to representations; that the computational instructions must also be stored, and the conversation veered at one point into the discussion of the lexicon containing representations of previous computations (we will return to this below). It was also noted that representations need not be stored, as they are also the output of a computation. The definition of a computation then fell out as any element that can have an input and an output. The structure of possible computations was not debated in detail (how to define a possible rule/constraint/system), and might be a fruitful area of discussion in the comment section below.
The afternoon consisted of the Make a claim and defend it session. Because this format was new I’d like to take a little time to discuss it before I get into the topics covered... Here, presentations were 5-10 minutes long, followed by 20-25 minute discussion sessions. I personally found this to be a great format that allowed for greater depth of analysis of the implications of each claim than is afforded in a more normal conference set-up. I do think that any relatively small and cohesive group would benefit form more meetings organized in this way (cohesive regarding the scope of underlying competencies of the participants). One difficulty with this format is a prolongation of the same phenomenon seen in more typical conferences, where each question period is monopolised by a few senior members of the group. It was suggested in the wrap-up meeting that next year there be more care taken to ensure that the discussion periods were more balanced, and that more students participated. In that meeting we started forcing more student participation, to great effect. If you threaten to sit and stare at them until they speak, they speak!
So, in this Make a claim and defend it afternoon session Heather Newell (UQAM) (me) and Guillaume Enguehard (Univ. Paris 7 / CNRS) started off the discussion with two presentations on the role of the morpho-syntax in the phonology. I argued that the Phase Impenetrability Condition cannot explain freezing effects in the phonology, and that we have to be careful about which operations we (can) take from the syntactic module. So, in effect, I argued that freezing effects in the phonology must be representational and not computational (without presenting any detailed account of how the representational account would function). Guillaume argued that Scheer’s account of Verner’s Law leads to a double representation of stress, which is undesirable. He proposed that Verner’s law can be accounted for by the insertion of an initial CV at the edge of morpheme boundaries. I think everyone agreed in this session that whatever generalizations stem from the morpho-syntax have to be outside of phonology, and that an investigation of these phenomenon will get us closer to a theory of what theoretical objects we need to account for in the phonology, even if we did not necessarily agree on all of the details of these analyses.
After the first break Sławaomir Zdziebko (Lublin), Joaquim Brandao de Carvalho (Paris 8), and Faith Chiu (UCL) & Typhanie Prince (Nantes) presented work that touched broadly on the problem of distinguishing between analyses that are computationally or representationally based within a single data set/phenomenon. Sławaomir offered a proposal that defines the boundary between suppletion and processes that are traditionally called ‘readjustment rules’. He proposed that the heads triggering readjustment rules are morphemes that have floating phonological structure that targets the closest possible host. This allows readjustment rules to be completely distinguished from allomorphy, bringing them into the realm of phonology; thus resolving the main criticism of readjustment rules in the literature. I noted in the discussion that it would be really nice to have independent evidence for these featural morphemes to really pit this analysis against a suppletion-only account of all morphological relations. Joaquim talked about the distinction between processes that are triggered in the morphology or in the phonology from a strict structuralist perspective, and proposed that morpheme structure constraints can be doubled in the phonology. This was, as it generally has been, quite controversial and led to a lively debate about if we want a grammar that effects a single surface pattern by more than one path. It was generally concluded that we don’t, but John Harris (UCL) brought up at the end of the discussion that there is psycholinguistic evidence that the same representation can be either stored or derived in the same language, giving us food for thought. Faith and Typhanie then brought in evidence from processing and aphasia that point towards the conclusion that syllable structure is stored in the lexicon. It was argued that aphasic patients with processing difficulties still have access to syllable structure, indicating that it is stored, in line with Government Phonology, and contra theories where syllabification is computational. It was noted by the audience that the complexity of the deficiencies exhibited by these kinds of subject may make it difficult to tease out exactly what the data mean here, but all agreed that this kind of work leads to interesting questions and results.
After the second break we had our final two presentations, which both discussed the features that specific representations have, and what they should be representations of. Ali Tifrit & Laurence Voeltzel (Nantes) discussed what evidence from metathesis tells us about the distinction between CVCV theory, and Government Phonology 2.0. Ali and Laurence argued that a CVCV phonology account of the data involves movement operations that target skeletal (CV) position, and that a GP2.0 account can do the same work in a purely representational manner. This led to a lively discussion in which Tobias pointed out that they had tried to apply his analysis to a different dialect, that he disagreed with this transposition since the data show that the pattern of metathesis is different in the dialect they analysed (and he offered an alternate analysis, subject to being more intimate with the data). Above all, in the end everybody involved agrees that CV units aren't displaced during a derivation. Eugeniusz Cyran (Lublin) offered various arguments supporting the proposal that melodic representations need to be substance-free (the relationship between phonetics and phonology is arbitrary) rather than phonetically-based; the latter view being supported by many phonologists working in element or feature based theories. The room agreed (perhaps unanimously?!). An interesting discussion followed, steered by Silke Hamann ( University of Amsterdam), regarding how to constrain and learn substance-free segmental structure. This conversation touched on what (or whether) the universal nature of phonological representations and computations is, and on how much overlap there is between phonological processing and general cognition. The importance of this topic was indicated by the fact that this debate was carried over into the summary meeting the following morning, but, of course, no specific conclusions on this topic were reached in the span of a single day.
In the end, two topics reigned in all of the discussions. First, the discussion of what a representation can look like, and secondly, the question of how to distinguish elements specific to the phonological module from elements stemming from other entities (general cognition, the lexicon, the morpho-syntax). Tobias, rightly, called this the Holy Grail problem (attributed to Ricardo Bermúdez-Otero (Manchester)).
Conclusions and consensus were found in the domain of computations; these are operations that have an input and an output. Questions and debate remained at the end of the day over what a representation is, and how to define the distinction between representations and computations. We cannot see this lack of ‘big movement/conclusions’ as a downside of the day. It makes me think of a conversation I had over lunch after the morning tutorial with Marc van Oostendorp and Cormac Anderson (Poznan) about a problem in linguistics, and scientific investigations in general, that I’m sure many of you have discussed before yourselves. It is clear that the academic system is set up in a way that all papers deemed publishable need to have a central claim that is in some way ‘big’, some ‘wow-factor’ if you will. This actively discourages the tinkering with and focusing on small details that would help us converge on more nuanced analyses, (besides perhaps squib format articles). I think the success of the first PTA is that we constituted a group that started from a basic underlying foundation (see the PTA manifesto) and started hacking away at the project of how and where the theoretical tinkering needs to be done within current instantiations of phonology. I look forward to the upcoming PTAs (the next ones are proposed to be attached to the next Going Romance and MfM meetings) and all of the big and small steps forward to come.