This file contains instructions for the steps we followed for the curve fitting analysis of intonation in the Greek in Contact project. The instructions below assume that a number of .wav and Praat TextGrid pairs have been created. This process has been applied to the processing of intonation in three types of utterances (declaratives, polar questions, and continuation rises) in six language varieties (Athenian Greek, Cretan Greek, Asia Minor Greek, Corfiot Greek, Cypriot Greek, Venetian Italian, and Turkish). You can read some of the results published at https://greekincontact.phon.ox.ac.uk/research. Note that separate scripts, not included in this document, are necessary to complete some of the steps below, which will be pointed out when they are mentioned. The commands included below have been created to run in a Linux environment but should also work in a Mac terminal. The main data of the project, as well as the scripts necessary to process the data have been deposited to https://doi.org/10.5287/bodleian:keKQbgg7Y.The way we speak is influenced by factors such as age, sex, where we grow up and social interactions over our lifetimes. Most people know that regional dialects may use different words (a 'bread roll' is a 'cob' in the East Midlands) or vary in individual sounds (Londoners will tend to pronounce 'think' as 'fink'), but are generally not aware of differences in intonation, the melody of speech. For example in Southern British English the sentence Jack did it, ending with the voice going down, typically signals a statement, but when the voice goes up it becomes a request for confirmation. But in Belfast English a rising pitch is used for both meanings. Therefore our speech patterns reveal not only what we want to put across, but also a speaker's language or region. Studying those differences has the power to uncover past and present ethnic interactions. In this project we analyse the selected speech melodies from regions where Greeks lived alongside speakers of other languages. Broadly, our main goal is to find out how the differences in the details of those melodies inform us about the nature of contact between a) Greek and Turkish and b) Greek and Italian populations. A sample of recordings we analysed shows that Greek dialects hailing from the Anatolian peninsula (modern Turkey) have Turkish traits, even though the people whose speech we examined are not in contact with Turks any longer. This is a promising finding, making a contemporary dialect a window to the past: it shows that the Greek and Turkish communities had not only co-existed, but they closely interacted. We will test this initial result on a larger corpus of Greek-Turkish speech from the Anatolian peninsula (continuing the research we started for this region), and Cyprus, to ensure statistical robustness of our result. Cypriot intonation patterns can throw light on the extent of Turkish-Greek interactions on the divided island. We will also analyse the intonation of Greeks from Crete and Corfu (where Italian speakers used to live), to discover what it reveals about past Greek-Italian interactions on those islands. This is our second goal i.e. to see how fine differences in speech melodies may have resulted from different lengths and types of population co-habitations. Our third goal is to determine how intonation continues to change over time after contact between the languages in question has ceased. To do so, we will carry out a longitudinal study of the Anatolian peninsula Greek to establish how the changing history of the contact between Greek and Turkish speakers has affected the intonation of this dialect over time. We are in a unique position to do so because we amassed a corpus of recordings of five generations of Greeks from Asia Minor and the Balkan peninsula (dating back to 1917), courtesy of several institutions and researchers. The value of our investigation lies in a novel approach to the study of language history and change. There are two aspects to this: first, we study a hundred years of historical influences on a language through the analysis of spoken records rather than the typical investigation of texts; second, we compare intonation instead of the more common investigations of vowels and consonants. The first has not been feasible in such depth of time due to the near-impossibility of collecting such corpora. The second has not been attempted within the traditional historical linguistic framework that relied on comparisons of individual speech sounds from audio recordings or written transcripts. Our method combines mathematical modelling, speech processing and theory of intonation in order to refine general linguistic theory, a combination not applied before. The insights gained from this project could then be applied to other languages, e.g. the influence of French on English in Canada or Spanish on English in Gibraltar, with the ultimate goal of understanding how languages change over time.

The .wav files included in the data for this project are short extracts from bigger recordings, small fragments of corpora from which they were extracted. They are not simply copies of corpora but our own processed files. All the data sources were digital, in a variety of formats (e.g., mp3, mp4 and .wav PCM; 2-channel or monophonic), bit rates or sampling rates (e.g., 44.1 kHz, 22.05 kHz, or 16 kHz). Additionally, some digital recordings made from ¼ inch tape, recorded at different tape speeds, required speeding up or slowing down by a factor of 2 or ½ to restore the correct original recording rate. A small number of such digitised tape recordings ran backwards on one channel, as the tape spool had originally been turned over for the second half of a monophonic recording, but it had been digitised as if it were a 2-track stereo recording. To permit for the subsequent functional data analysis steps to be performed as batch computations, we converted all the recordings to 16 kHz, monophonic, uncompressed PCM .wav audio files. Each of the resulting .wav files included here contain only the “Region of Interest” interval from each recording: this interval is a stretch from one to three syllables, depending on the variety and stress position (antepenultimate, penultimate, or final).