SlideShare a Scribd company logo
1 of 98
Download to read offline
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
Contact Effects in Phonology: A Case-study of Finno-Romani
Index
Abstract ............................................................................................................................................ 2
1. Introduction ................................................................................................................................. 2
2. Background .................................................................................................................................. 5
2.1. Sociolinguistic explanations ................................................................................................ 5
2.2. Structural explanations ........................................................................................................ 9
2.3. Typological explanations ................................................................................................... 12
2.4. Summary ............................................................................................................................ 13
3. Method ....................................................................................................................................... 14
3.1. Data .................................................................................................................................... 15
3.2. Procedure ........................................................................................................................... 17
4. Preliminary data analysis ........................................................................................................... 20
4.1. Cases which might be both loanwords and inherited words .............................................. 21
4.2. Cases with semantic shift ................................................................................................... 22
4.3. Cases with different cognates ............................................................................................ 24
4.4. Morphology ....................................................................................................................... 25
4.5. Other considerations .......................................................................................................... 25
5. Data analysis .............................................................................................................................. 26
5.1. Analysis of correspondences with vowels ......................................................................... 27
5.2. Discussion of vowel inventories ........................................................................................ 33
5.3. Analysis of correspondences with consonants ................................................................... 36
5.3.1. Sonorants .................................................................................................................... 36
5.3.2. Obstruents .................................................................................................................. 41
5.4. Discussion of consonant inventories .................................................................................. 51
5.5. Stress and length ................................................................................................................ 58
5.5.1. Stress .......................................................................................................................... 58
5.5.2. Length in vowels ........................................................................................................ 59
5.5.3. Length in consonants ................................................................................................. 61
6. Discussion .................................................................................................................................. 63
7. Conclusion ................................................................................................................................. 70
References ..................................................................................................................................... 71
Appendix 1 – Sorting of wordlists ................................................................................................. 74
Appendix 2 – Cognates used in the analysis ................................................................................. 82
Appendix 3 – Correspondence sets ................................................................................................ 85
Appendix 4 – Protoforms ............................................................................................................... 96
1
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
Abstract
This paper investigates phonological contact effects through a case study of Finno-Romani.
Existing research on the topic is presented, covering sociolinguistic, structural and typological
studies. The development of Finno-Romani since its contact with Finnish is examined using the
comparative method, and the changes are discussed in relation to the sound system of Finnish. I find
that many changes are predictable from a combination of knowledge of the sociolinguistic situation
and analysis of the sounds of the involved languages. In addition to this, I also find that certain
changes in Finno-Romani cannot be explained solely with contact with Finnish, but that these
changes have nevertheless happened because of it. I discuss two different views of phonological
change and how they are applicable to the case of Finno-Romani, and find that although the two are
not immediately reconcilable, they each contribute important information. I conclude that more
studies with a structural focus are needed in order to better understand the mechanics of
phonological change in contact situations.
1. Introduction
In this thesis I systematically examine the process of phonetic and phonological change in a
language in close contact with another language. The goal is to produce a thorough case-study of
the phonetic and phonological aspects of language contact, and to propose future ways to study this
phenomenon. The case-study in question is of Finno-Romani which has been in very close contact
with Swedish and Finnish for a long time, and is reported to have changed significantly as a result
of this contact, including its phonology (Granqvist 2002; Granqvist 1999a; Granqvist 1999b).
Romani is well suited for this kind of study because its various branches have been in contact with
many different languages. For this reason, there are rich grounds for future comparisons. Finno-
Romani in particular is suited for this explorative case-study as its geographical placement means
that it has a much more limited range of possible contact languages than for instance some of the
Romani dialects of Central Europe. When reconstructing the development of Finno-Romani's sound
system since its split from the other Romani dialects, I compare the changes in Finno-Romani with
the phonological systems in its latest contact language, Finnish. This is in order to determine which
of the possible contact-induced changes have happened and which have not, and what this might
reveal about the mechanics of language change in contact situations.
There are many different aspects of language contact and many different ways of studying
2
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
them (sociolinguistic studies, studies of politically motivated language change, studies of mixed
languages and studies of language attrition to name but a few), and all of these aspects are
obviously interconnected and can each provide a deeper understanding of the others. In section 2 I
provide a general overview of the suggested explanations of contact-induced language change that
seem most relevant to the present study. However, as this subject field is so enormous, I focus in the
main analysis almost exclusively on the structural aspects of language change and what these might
reveal about the general workings of human languages.
Languages are in close contact with other languages all over the world, and the situation in
which they start to affect one another can reveal much about the nature of human language. From a
structural point of view, any situation in which two language systems are in contact with each other
can reveal something crucial about both the languages involved and the human language capacity.
Or, as Matras (2009: 3) puts it, it can help us gain “an understanding of the inner functions and the
inner structure of 'grammar' and the language faculty itself”.
For this reason, as well as for many others, the field of contact linguistics is large and
growing (Matras 2009: 3), but to the best of my knowledge no studies like the present one have
been carried out: A study in which a careful mapping of the phonetic and phonological language
history is compared to that of the given language's contact language(s). Two studies examine
phonetic and phonological effects of language contact: Shaw & Balusu (2010) study changes in
Japanese fricatives caused by contact with English, and Lleó, Cortès & Benet (2008) deal with
changes in vowels in Catalan caused by contact with Spanish. Both do so by choosing a specific
sound type and examining only that. In the case of Lleó, Cortès & Benet 2008, it is not even clear
what the motivation for choosing those specific sound types is. Inspired by the point made by
Thomason (2001: 93) that it is important to study the whole of a system when determining whether
a change is contact-induced or not, I therefore examine the whole of the segmental inventories
involved, as well as stress patterns and length in vowels and consonants. Ideally, the scope would be
even broader and include not only phonotactics, but also morphological, syntactic and
sociolinguistic factors as well, in addition to more precise phonetic measurements. This would
provide a larger context for the sound changes in language contact studies. However, given the
limited space available for this study, I have deemed it more useful to pick out a single, clearly
demarcated subject of analysis instead of including bits and pieces from all domains.
The term 'language contact' covers a broad spectrum of different kinds and levels of contact.
My focus in this paper is on situations in which one language has caused considerable change in
3
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
another but not (at least not yet) caused actual language death, and in which the changed language
can still be recognised as the original language. That is, I will not focus on actual mixed languages
or creoles, nor on language attrition or substratum effects, only on intensive contact in a situation of
bilingualism.
As already mentioned, my focus within the area of language contact is phonetics and
phonology. There are several reasons for this. First and foremost, it is a field of study which –
compared to contact morphology and syntax – has been neglected in recent years. Second, given
that phonetics and subsequently phonology work within the boundaries of the human articulatory
capacity, it offers a more finite set of variations than morphology and syntax, which are limited
mainly by cognition. This limitation makes it easier to gain a good understanding of which contact
effects are more likely to occur as many can be ruled out from the very beginning.
Given the focus on the structural aspects and the declared goal of investigating the human
language faculty through it, I will also touch (however briefly) upon the psycholinguistic aspects of
bilingualism and multilingualism, as it is neither entirely possible nor indeed desirable to separate
the synchronic aspects of language change from the diachronic aspects. In this I am inspired by
Matras (2009: 2-6), whose declared goal it is to incorporate synchronic as well as diachronic studies
into a coherent, functionalist model of communication which also takes into account “inner
coherence of language systems” as well as models of bilingual language processing, typological
findings and contact linguistics. While the construction of such a model is very much outside the
scope of this study, I nevertheless present my analysis in a way that might later be useful in the
discussion of such models. This is clearly an area of study which is nowhere near exhaustion and
one which I hope to contribute to.
The structure of this paper is as follows: In section 2 I give a more detailed presentation of
the work already done in the field of contact linguistics in general and in phonetic and phonological
aspects of language contact in particular. I then explain my choice of method and the data used in
the analysis (section 3), and then move on to the analysis of my chosen case: Finno-Romani
(section 4). This is followed by an actual analysis of the sound changes that have taken place in
Finno-Romani since its arrival in Finland, and these are compared with Finnish throughout the
analysis (section 5). I round off this presentation with a discussion of my findings and the
theoretical consequences they have for the study of phonetics and phonology in contact situations
(section 6) and I conclude the paper in section 7.
4
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
2. Background
I now present the research area of contact-induced sound change. In order to make the presentation
more structured, I split the topic into three different branches, following Matras 2007. Each of these
branches will be presented and related to the subject of the case-study presented in section 5.
In an overview of proposed explanations of structural borrowing, Matras (2007: 34)
identifies three different directions that explanations of contact-induced change might take: (1)
Explanations that are concerned with “the degree of borrowing as related to the intensity of
exposure to the contact language” (from here called 'sociolinguistic explanations'), (2) Explanations
that are concerned with “the outcome of language contact as a product of the structural similarities
and differences (congruence) among the languages concerned” (called 'structural explanations'), and
(3) Explanations that are concerned with borrowability as a product of inherent traits of a given
category (called 'typological explanations').
As all three aspects (sociolinguistic, structural and typological) are obviously relevant to an
explanation of a contact situation and its effects, I will not in the following limit myself to dealing
with only one of them. In fact, they are not entirely separable anyway. The main analysis in this
study, however, is designed to cast light on the process from the point of view of the second type of
explanations – the structural. In order to give the reader a solid context for the structural analysis, I
also give an overview of the first (sociolinguistic explanations) and third (typological explanations)
types of explanations, and provide the most basic information needed to understand how Finno-
Romani might be explained according to these.
2.1. Sociolinguistic explanations
Without doubt the most influential work concerned with the degree of contact and its consequence
for the degree of borrowing, is the monograph by Thomason & Kaufman (1988). As is very clearly
stated, Thomason & Kaufman hold the view that, while purely linguistic considerations are
relevant, “it is the sociolinguistic history of the speakers, and not the structure of their language,
that is the primary determinant of the linguistic outcome of language contact” (Thomason &
Kaufman 1988: 35). While it has been suggested that this is not quite right (Treffers-Daller 1999),
the view has also been defended from many different sides (see for instance Beeching 1999;
Poplack & Meechan 1999; Sebba 1999; Thomason 1999; Winford 1999). To my knowledge nobody
disputes that the degree and type of contact is indeed very relevant to the linguistic outcome of the
5
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
contact situation. In the frame-work of Thomason & Kaufman (1988: 50), the sociolinguistic
circumstances of Finno-Romani can be characterised as being a case of language maintenance (as
opposed to language shift) as Finno-Romani is not quite in the process of being abandoned by their
speakers (see Vuorela & Borin 1998 for a more nuanced view on this though). In addition to this, it
is a case with “intensive contact, including much bilingualism among borrowing-language speakers
over a long period of time” (Thomason & Kaufman 1988: 50). Indeed, speakers of Finno-Romani
are probably all dominant in Finnish. According to the model, this leads to a high degree of lexical
borrowing, and moderate to heavy structural borrowing, especially at the phonological domain. This
is to some extent true, although the lexical borrowing from Finnish into Finno-Romani is
surprisingly limited (for more on this, see section 3.2). Finally, Finno-Romani can be argued to be
under a fairly strong cultural pressure from the surrounding Finnish-speaking society (see however
again Vuorela & Borin 1998 for a more nuanced presentation of the situation), which is predicted to
lead to massive grammatical replacement. This is also true, depending on the exact definition of
'massive'. The model thus seems to predict the results of the contact quite well based on the
sociolinguistic parameters of the situation.
In a similar tradition, but within a more speaker and communication oriented model, Matras
2009 proposes a functionalist view of communication in contact situations which explains contact-
induced language change with the bilingual speaker as the central actor. In this model, change in
contact situations happens to balance two opposing tendencies: The tendency to remain loyal to a
specific repertoire in specific contexts (that is – to keep the acquired languages strictly separated
and to keep each language linked to its specific social contexts) and the tendency to exploit the full
expressive potential of the whole of the available repertoire (that is – to mix the acquired languages
in order to be able to choose the most fitting trait in the situation, no matter which language specific
items belong to). The way to balance these things is to reduce the amount of choices the bilingual
has to make between the two opposing tendencies. This balance can be achieved for the speaker by
making the involved languages more similar to each other. If one language borrows a word from the
other, it becomes unnecessary for the bilingual speaker to make a decision of whether or not to use
the given word in a context that invites loyalty to the recipient language. At the same time, the full
expressive potential is freely available.
When it comes to phonology in a contact situation specifically, Matras (2009: 221) notes
that certain aspects are notably different from other grammatical domains such as morphology and
syntax. Firstly, sound material falls somewhere between matter (content) and pattern (structure; for
6
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
a more thorough discussion of these terms and their roles in language change, see Sakel 2007).
Sounds obviously have a clear and audible content (and therefore matter), but at the same time
sounds on their own cannot carry actual meaning in the way that for instance morphemes can.
Therefore, the borrowing of sound patterns is not quite comparable to the borrowing of grammatical
structures. Secondly, physiology restrains phonetics and phonology, but not morphology and syntax.
There are therefore limits to the speakers' choice of whether or not to adopt new sounds from a
contact language, especially if the speaker is an adult.
Matras (2009) claims that all levels of phonetics and phonology may be borrowed, from
phonetic features to phonemes to more prosodic features. Based on various degrees of contact (and
in this his model is similar to that of Thomason & Kaufman 1988), he then lists four different types
of processes leading to contact-induced phonological change. These four types (called A, B, C and
D) cover both changes in a speaker's native language and in the contact language as spoken by the
bilingual non-native speaker. In type “A” changes happen to elements borrowed into a language and
not to the language that borrowed them. This is the kind of change that Singh (1996) calls
“adaptation” and in which the phonology of borrowed words are adapted to fit the phonology of the
recipient language. In Matras' speaker-centered model, this reflects a situation in which the speakers
wish to remain loyal to their native language (while obviously still exploiting the potential for
expression, given the fact that the borrowing has happened at all).
In the type called “B”, there is also borrowing of words from a contact language into the
recipient language, but no phonological adaptation is taking place. Rather, the borrowed words are
allowed to keep their original shape leading to an enrichment of the phonology of the recipient
language. This is what Singh (1996) calls “incorporation”, and this requires a degree of competence
in the bilingual speakers which is not necessary for adaptation. In Matras' model, this reflects a
situation in which the speakers are less concerned with staying loyal to their native language, and
they are more concerned with allowing borrowed words to keep their foreign traits. This is
generally due to the source language having a higher status than the recipient language. Both type A
and type B describe situations in which individual borrowed words are affected and not the whole
phonological system.
In the type called “C”, there is widespread bilingualism, and speakers are uncomfortable
maintaining two separate phonological systems. The solution to this for the speakers is to speak the
contact language with something close to, or significantly influenced by, the native language's
phonology. This is frequently described as 'substrate effects' and in individuals the process is
7
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
commonly known simply as 'accent': The speakers are competent in both languages but find it
easier to use the same phonological system (or at least parts of it) in both. In Matras' model this
helps speakers avoid the choice between which phonological system is most fitting in a given
context. One might argue that if the language spoken has to be decided upon, then the choice of
phonology is a simple consequence of this choice and should therefore not offer any dilemma.
Matras does not specifically comment on this, but, as already mentioned, it is noted that there are
physiological limitations as well as communicative ones when it comes to phonology. I find it likely
that physiological ability is more relevant to the phonological aspect than the competing forces of
loyalty and communicative efficiency. Type C is similar to type A in that there is a transfer of the
native language's features onto the foreign material, and not the other way around.
In the type called “D”, something similar to type C happens, but in this case the contact
language is dominant enough (that is, it has high status and there are many and competent
bilinguals), that it is the phonology of the native language ('native' in this case being a somewhat
dubious label as it might not always be the speaker's first language) which is adapted to fit the
phonology of the contact language. Type D is similar to type B in that native material adapts to fit
the foreign sound patterns. Type D is similar to type C in that speakers in both of these types try to
avoid the selection of sounds from two distinct inventory sets and therefore end up using just one
set.
There is no claim that the four types are meant as each describing only one specific and
distinct type of language contact, as there is overlap between the types. Furthermore, various types
can often be found in the same contact situation. The main claim is that speakers in contact
situations constantly navigate the opposing pressures of loyalty to one given language and the
practical urge not to have to choose between two separate sound systems. The result is that speakers
find themselves somewhere on a continuum which goes from having two completely separate
systems and having the systems overlap and essentially become one. The placement on this
continuum correlates with the degree of bilingualism found in the community: The higher the level
of bilingualism, the more likely a convergence of the involved systems is found.
While it is outside of the scope of this study to discuss sociolinguistic factors thoroughly, I
will in this section - drawing heavily on work done by Vuorela & Borin 1998 - briefly summarise
what is known about the social and political conditions of Finno-Romani in present and past. This is
intended to give the reader an idea about which sort of language contact is being dealt with here,
and where contact effects might have come from.
8
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
The Roma-speakers of Finland are descendants of a group of Roma who came to Sweden in
the early 16th
century, after having split from the other Roma groups (which are now spread across
Europe). They first appeared in Finland (which was a part of Sweden until 1809) in the late 16th
century (Vuorela & Borin 1998: 52). The Roma continued to come to Finland from Sweden until
1809, when Finland became Russian, and the Roma residing in Finland at that time are thought to
be the ancestors of the modern day Finnish Roma (Vuorela & Borin 1998: 53). To this day, many
Finnish Roma (an estimated third of the total 9000) live partly in Finland and partly in Sweden
(Vuorela & Borin 1998: 55). During the centuries since the Roma's arrival in Finland, they have
been singled out as a group with various negative traits by authorities (Vuorela & Borin 1998: 54).
and for this reason Finno-Romani has always been a low-status language. All Finnish Roma also
speak Finnish and for many of them it is their first and perhaps even only language. Nevertheless,
Romani is kept alive as an in-group language which is learnt in late childhood as a part of the
adoption of Roma cultural behaviour (Vuorela & Borin 1998: 60). In 1998 Finland signed the
European Charter for Regional and Minority Languages and identified Romani as a non-territorial
minority language of Finland (Ministry of Social Affairs and Health of Finland 2004: 22). This
however does not mean that the Roma (and in extension their language) generally enjoy a high
status. All speakers of Finno-Romani are bilingual, and as already mentioned, Romani is in most
families learned in late childhood. Romani is therefore the second language of the younger
generations with Finnish being the first. According to the models of both Thomason & Kaufman
(1988) and Matras (2009), this puts the contact situation between Finnish and Romani in the groups
that is expected to show the most contact effects. According to Matras' model, given this degree of
contact, the two phonological systems are likely to have more or less converged into one. Whether
or not this expectation holds true will be discussed further in section 6.
2.2. Structural explanations
When it comes to the more structurally-oriented explanations of the processes of contact-induced
phonological change, there is to my knowledge no coherent model of what might be the expected
outcome when two languages are in close contact. However, there are several smaller contributions
to the field, and I will present some of them here.
Still within the framework of the model described above, Matras (2009) describes various
types of phonological contact effects based primarily on the empirical work described in (Matras
9
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
2007). As these various effects do not seem to be grouped or in any hierarchy, they will simply be
presented one by one.
One process that is commonly found in situations with a great amount of contact, is that
sounds in loanwords are reinterpreted as native sounds. Matras (2009: 226) notes the following of
the process: “Speakers perceive similarities between a sound X in one language, and a sound Y in
another. These similarities usually derive from shared features in the position of articulation of the
two sounds, sometimes also in the mode of articulation. On this basis, one sound is allowed to
represent the other.” In types A and C this process happens with the native language as the template
and the borrowed words as something which is mapped onto it. This process is called
'approximation'.
Another process is the loss of features, and this process leads to the omission of one or more
features in a phoneme in a loanword. The omission is typically due to the feature being difficult for
speakers of the recipient language to pronounce, and omissions are therefore unlikely to have
counterparts in the recipient language. This might happen not only to single features but also to
phonemes or specific phoneme combinations such as consonant clusters, which might be reduced in
loanwords. In some cases these reductions of features or phonemes may result in a disruption of the
general phonological system, as some distinctions might be lost. This might lead to words which
were minimal pairs in the source language to merge in the recipient language.
Related to feature loss is phoneme substitution. In this process, phonemes in loanwords are
substituted by phonemes in the recipient language that share enough features with them, so that they
seem somewhat similar. According to Matras (2009: 227), this is a very common process, as it is
very likely that a given pair of languages share features of a given phoneme but a lot more unlikely
that they share the exact same combination of features in the same phoneme.
Syllable structure and stress systems too are susceptible to changes when loanwords are
incorporated into a recipient language, although this does not always happen. Even when no actual
changes in the phonological system happen, there might be a frequency effect when a certain
phoneme with a somewhat peripheral function in the recipient language is given a more prominent
role due to its occurrence in many loanwords. It might also happen that loanwords are allowed a life
outside the general phonological rules of the recipient language. In contrast to this it might also
happen that an extensive borrowing leads to a deeper change in the phonology than merely an
enrichment of the inventory. If, for instance, a sufficiently large amount of loanwords are borrowed
in which a given phoneme type is found in positions where it is not usually found in the recipient
10
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
language, then the established system might adjust. In cases with even more effect on the recipient
language's phonology (“convergence”, Matras 2009: 229), the similarities between the languages
involved increase. This might lead to enrichment of one of the inventories, but also loss of
phonemes in one of the languages. A very frequent stating point of convergence is the borrowing of
an allophonic variation and/or the rules for its distribution.
In direct relation to the notion that phonological convergence starts with the borrowing of
allophonic variations or distributions rather than of new sounds, Shaw & Balusu (2010) present a
study of changes found in Japanese due to contact with English. Their proposal is that new
phonological distinctions that are found in loans, are parasitic on existing phonetic (but non-
phonemic) differences found in the receiving language. The example studied by Shaw & Balusu
(2010) is the phonological contrast between [ti] and [tʃi] which is found in Japanese in loanwords
from English. Japanese does have the contrast [ti] – [tʃi], but only as a prosodically conditioned
difference: [tʃi] appears in prosodically strong environment such as heads of feet and accented
syllables, while [ti] appears elsewhere. The phonological contrast in the English loanwords thus has
a ready phonetic continuum to map unto, and it is the suggestion of Shaw & Balusu (2010: 157) that
this is a prerequisite for the preservation of the contrast in the loanwords. The suggestion is
strengthened by the reported fact that most contrasts not found in the native phonology are simply
neutralised in English loanwords in Japanese.
Singh (1996) combines structural explanations with typological ones to explain why certain
features will be more easily borrowed than others. Citing The Universal Syllabic Template (which
states that the nucleus of a syllable is the peak of sonority and that going out from this peak (to both
sides) the sonority will be steadily falling), he states that it is easy to borrow something that fits a
general principle (such as the Universal Syllable Template) even if the recipient language originally
has a gap in that place. As an example he mentions that English does not have the consonant
combination [vl] in syllable onset even though this fits the Universal Syllable Template. However,
the fact that it does fit the Universal Syllable Template, makes it easy for English speakers to
pronounce borrowed words from Russian such as 'Vladimir', and this fact will make words with
these phoneme sequences more likely to be borrowed unchanged into English than others.
Lleó, Cortès & Benet (2008) set up two different sets of expectations to the outcome of a
study of Catalan vowels during contact with Spanish. One of the sets of expectations is based on a
general typological basis while the other is based on a comparison of Spanish and Catalan
phonologies. While the results are not entirely clear-cut, the comparison of the specific phonologies
11
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
is deemed by the authors to be the best predictor of the outcome. Catalan has a vowel system with
four mid-vowels: /e/, /o/, /ɛ/ and /ɔ/ whereas Spanish only has /e/ and /o/. As the Spanish system is
less complex, it is expected that Catalan in contact with Spanish will have a tendency to merge /e/
with /ɛ/ and /o/ with /ɔ/ (cf. Matras' description of phoneme loss in section 2.2). The authors wonder
which of the two original mid-vowel sets ([e] – [o] and [ɛ] - [ɔ]) will be the realisation of this
merged phone in Catalan. In order to investigate this, they make recordings of three generations of
speakers of Catalan living in two different parts of Barcelona with different proportions of Catalan
speakers. It turns out that in speakers who displays most contact effects, the [e] - [o] vowel set is the
one primarily found in Catalan as the realisations of the merged phonemes. Lleó, Cortès & Benet
(2008) do not really explain why they chose to look at vowels rather than something else, or how
they started suspecting that there might be something to look for. I find that this lack of insight into
the whole of the phonological systems makes the study a little hard to assess.
As structural explanations of contact-induced change are the main focus of sections 3 - 6 and
as they will be treated there, I will not elaborate on the phonological structure of Finno-Romani or
Finnish here.
2.3. Typological explanations
Explanations of phonological contact-induced changes based on typological factors are relatively
rare, and the reason seems to be that there simply are not very many: “it does not seem possible at
this stage to point to any position within the phonological system (e.g. certain articulatory modes or
positions, marked features, etc.) as being particularly prone to contact-induced change. It seems that
the details of phonological change are entirely a product of the relations among the two systems –
or congruence – and any statistics of change are likely to simply reflect the mere likelihood of the
two phoneme systems in contact to share certain phonemes, and to differ with respect of others.”
(Matras 2007: 39). Matras (2009: 228) does add that new consonants are more likely to be adopted
than new vowels, but this is attributed to the fact that consonant inventories are generally larger
than vowel inventories and that there is therefore a higher potential among the consonants of not
overlapping between the involved languages. This entails a higher pressure on the consonant
systems to adjust than the vowel systems. Also, prosodic features seem to be more susceptible to
borrowing than segmental phonology, which is attributed partly to the somewhat peripheral role
prosody has to meaning and partly to “the proven neurophysical separation between prosody and
12
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
other aspects of speech production” (Matras 2009: 233). Here it is quite relevant to mention that
Matras (2009: 231) does not seem to count stress as a prosodic feature stating “prosody seems to be
more prone to cross-linguistic replication in contact situations than segmental phonology, with
stress figuring in-between the two”.
It has been proposed that markedness is highly relevant for the likelihood a trait has of being
transferred. Lleó, Cortès & Benet (2008) present a “Jakobsonian perspective” on markedness (for
more on this, see Jakobson 1968) in which “marked entities presuppose unmarked ones in a
typological, diachronic and acquisition sense” (Lleó, Cortès & Benet 2008: 188), meaning that for a
language to have a marked trait, it must also have the unmarked corresponding version of it. It is
never made entirely clear by Lleó, Cortès & Benet (2008) what exactly one should understand by
'corresponding'. In the case of phonology it seems clear enough that a voiced stop can be said to be
more marked than an unvoiced one with the same place of articulation, but how far the notion of
'correspondence' can be taken is unclear. Is an unvoiced fricative in any way 'corresponding' to an
unvoiced stop, for instance?
Another factor which Lleó, Cortès & Benet (2008: 188) propose as relevant to borrowability
is complexity. 'Complexity' is meant as a psycholinguistic notion and is specifically linked to
allophony. The more allophones a phoneme has, the more complex it will be, as it will cost the
learner more work: There is more decoding to be done, and also more rules to learn in order to
produce the right phone in the right environment. Although it is never overtly stated, the assumption
seems to be that a complex trait might be harder to borrow than a simple one. This assumes that
phonemes are borrowed as a complete packet with all the allophones of the source language.
Finally, Lleó, Cortès & Benet (2008: 188) also suggest that traits which are frequent and
dominating in the source language are more easily transferred to another language. As Lleó, Cortès
& Benet (2008: 188) point out, this should not be confused with frequency in all the languages of
the world as that would be a matter of markedness rather than frequency. This suggestion is not
really followed up in the discussion of the results, but the overall conclusion is that a comparison of
the involved phonological systems (as described in section 2.2) is a better predictor of the outcome
of language contact than any general, inherent traits of the constituents of the phonologies.
2.4. Summary
In conclusion, sociolinguistic explanations of situations of language contact can successfully predict
the degree of contact effects found in the affected language(s). However, if one wants to know more
13
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
about the contact effects than merely their degree, other explanations and models are necessary.
While typological explanations have limited success in predicting the outcome of language contact,
more structural approaches do seem capable of predicting not only the degree but also more
specifically the type of contact effects. Still, there are limited resources on this in the existing
literature. Shaw & Balusu (2010) and Lleó, Cortès & Benet (2008) each cover only small corners of
the phonological systems of Japanese and Catalan respectively, and while Matras (2009) does list
many possible effects, he does not really comment on how these might be interrelated in one
phonological system. The overall focus missing thus seems to be a cohesive investigation of whole,
specific cases.
In order to at least begin to fill this gap, I now turn to an analysis of Finno-Romani and the
phonetic and phonological changes which have happened to it since it became a separate dialect
branch.
Finno-Romani is very useful for investigations of this sort. Firstly, various branches of
Romani have been in contact with so many different languages that there are very rich possibilities
for further comparisons. These comparisons might be studies of how similar languages behave in
contact with different languages – something that might give a greater understanding of what
language is and needs in order to be a language. This makes Romani in general uniquely suited to
this sort of study. Secondly, Finno-Romani is a collection of very similar but not quite identical
dialect variations. It is found in a relatively isolated part of Europe as Finland is geographically at
the fringe of the European landmass and also to a great extent surrounded by water, and addition to
this, speakers of Finno-Romani are rarely in contact with speakers of other Romani varieties. These
circumstances makes it, in a way, simply easier to work with than many other Romani dialects
spoken in for example Central Europe because there are limited sources of contact. I therefore deem
Finno-Romani a good starting point, and in the following I outline my approach to its investigation.
3. Method
In this section I describe the data I use and the data-base they are collected from (section 3.1) and
outline the procedure of the analysis to be described in sections 4 and 5 which is the basis of the
results summarised and discussed in 6 (3.2). This section also includes considerations about
possible but rejected procedures and the evolution of the remaining one, as it has changed a great
deal since the beginning of this project, and with good reason. I briefly describe the reasons for
14
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
these changes and why I find the present approach the most reasonable.
3.1. Data
The data used in this thesis is all collected from the Romani Morpho-Syntax database (RMS-
database, http://romani.humanities.manchester.ac.uk/rms/), (see Matras, White & Elšik 2009 for a
more detailed description). The RMS-database is a publicly accessible online collection of sound
files and transcriptions of speakers of various Romani-dialects reading aloud wordlists, translating
sentences into Romani, and engaging in narration. In addition to being simply a collection, the
database also makes various modes of cross-dialectic search available and is thus a very useful tool
in many different kinds of studies involving Romani (see for instance Elšik & Matras 2006) who
used it to compare morphological and syntactic changes found in the various dialects with the
languages they had each been in contact with). In most cases, the RMS-database does not give
information about their transcribers, but they are described as being mainly “graduate students
specialising in Romani linguistics” (Matras, White & Elšik 2009: 11). It is further noted that a great
amount of work went into assuring that the transcribers followed the same standards and
conventions for the transcription. It is also mentioned that the collection in the RMS-database was
gathered using an international network. I take this to mean that the transcribers working on Finno-
Romani were probably living in Finland as well and presumably native speakers of Finnish (which
they could be assumed to be even if they happened to be Finnish Roma, as mentioned in section
2.1). The transcription conventions used in the RMS-database are based on a standard transcription
consensus of Romani, which I have “translated” into broad IPA using both the sound files attached
to each word in the database and the existing transcriptions in the database. I found this the best
approach as I am well aware that there are important differences between various phonemes which I
being a Dane - albeit a linguistically trained one – would be hard pressed to notice, but which are
easily detected by native speakers of Finnish, especially in the realm of allophones. As several
sources note that Finno-Romani sounds a lot like Finnish (Granqvist 2002; Vuorela & Borin 1998),
it seemed the most prudent not to dismiss distinctions made by speakers of Finnish even in cases
when I myself could not hear it. In a very few cases I have deviated from the existing transcriptions,
as differences in the existing transcription and my own interpretation of the pronunciation seemed
large enough to justify it. These cases are discussed in the analysis. A last note on the transcription
is that as stress almost always falls on the first syllable in Finno-Romani, stress is marked only
15
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
when this is not the case.
When it comes to Finno-Romani, the database provides word lists from six different
speakers. Two of these are noted as speaking West-Finnish Romani (These are referred to as FIN-
005 and FIN-011) and the remaining four East-Finnish (Referred to as FIN-002, FIN-006, FIN-008
and FIN-012). One of the West-Finnish lists (FIN-011) is very incomplete and has been excluded
from the analysis for this reason. One of the East-Finnish lists (FIN-002) does not have
corresponding sound files which makes it impossible to verify the transcription given, but as the
transcriptions of the other dialects are generally good, I have chosen to include this sample. The
sound files of FIN-008 have a lot of background noise which makes it virtually impossible to judge
the transcription of especially fricatives and affricates, but this sample has been included in the
analysis too for the same reason that FIN-002 has. In the following, the different samples will be
treated as different dialects although there is only one speaker for each sample. It is quite possible
that others factors such as speaker age and the register used have a greater influence on the data
than the geographical placement and dialect variety does. However, as I am investigating different
outcomes of Finno-Romani in contact with Finnish, it is not so very important which of the factors
might be most influential.
In addition to the samples of Finnish Romani, I also use the two Sinti dialects available in
the database: One currently spoken in Romania (RO-022), which has until very recently been
spoken in Germany, and another dialect currently spoken in Austria (AT-001x). These are assumed
to belong to the group called German Sinti by Matras (2002: 9). I include these samples as Sinti is
closely related to Finno-Romani. Together they form the subgroup Northwestern Romani and they
have many lexical parallels (Boretzky 2012), which makes them suitable for an analysis and
reconstruction like the one described in section 3.2.
Using a resource like the RMS-database has a number of advantages. Most notably, it saves
the researcher doing an analysis very much time, energy and money not having to collect their own
data. In a collection like this, much work has also been done so thoroughly and with so competent
people involved that the quality of the data is probably better than what one person could collect
within a reasonable time-span. That said, there are of course also disadvantages: Mainly, that one
does not get to choose which information is relevant. I would have preferred to know if all the
samples were indeed transcribed by native speakers of Finnish as I have assumed, and also what age
the informants had and what their background was. I find that the advantages are greater than the
disadvantages, and so I judge it reasonable to use the RMS-database for this investigation.
16
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
3.2. Procedure
I now describe the procedure I use to get the results presented in this project. Briefly stated, the goal
is to determine which sound changes have happened in Finno-Romani since its split from Sinti and
other related varieties, and consequent close contact with (Swedish and) Finnish. These changes are
then compared to the Finnish sounds and sound structures in an attempt to determine whether or not
and to what extent the changes are due to contact with Finnish. The procedure has changed
somewhat in the course of the project, and as these changes are very much a part of the process of
working with the data, and as they are important to know in order to understand the reasoning
behind my method, I will describe those as well.
Using the data from the RMS-database, the goal is to determine the course of phonetic and
phonological changes in Finno-Romani and whether or not these can be ascribed to contact effects.
The first part of the procedure is to determine which words in the word-lists are inherited
and which ones are borrowed. The inherited words are needed for input to the comparative method
(described below) and this is made difficult if recently borrowed material is used, so this distinction
is quite relevant.
Originally, the intention was to also use the borrowed words because I assumed that most of
them would be from Finnish. The idea was to examine these loanwords in order to make a
description of the phonetic and phonological systems they represented when they were borrowed
into Romani. These systems were then expected to be the systems of Finnish at the time of the
adoption. While the form that was borrowed into Romani might very well have been adapted
already at the point of adoption, this was not considered to be a problem as I was happy to
investigate the part of the process taking place after the adoption. The process of adopting
loanwords is a highly interesting topic, but it is more easily examined in studies with a more
synchronic approach such as Lleó, Cortès & Benet (2008) and Shaw & Balusu (2010). I considered
the diachronic perspective of this study to be much more appropriate to investigate the process
happening after the adaptation (although this approach did assume that the adaptation of the words
to the Romani phonology was not complete, and that there would be structural differences between
the loanwords and the inherited words). My intention was then to compare the description of the
phonetic and phonological properties of the loanwords to the phonetic and phonological properties
of Romani and also to the changes which have happened in Finno-Romani since then. It became
apparent that this method would not yield any results when the data sets made it clear that there are
in fact very few loanwords from Finnish in Finno-Romani. The reason seems to be that Romani has
17
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
been used as a secret language by the Finnish Roma (Vuorela & Borin 1998: 73), and this would
obviously be less effective if too many of the words were recognisable to Finnish speakers.
While this plan was then abandoned, I still needed to know what Romani had looked like
immediately before contact with Finnish (or at least as close to 'immediately before' as it was
possible to get) and which sound changes had happened since then. Inherited words have been
separated from loanwords by a combination of comparing them with the other Romani dialects in
the RMS-database and finding etymological information in the relevant literature (mainly Boretzky
2012; Valtonen 1972 and Matras 2010). If a word appears to have clear cognates in other, more
remotely related dialects attested in the RMS-database, it is assumed to be an inherited word. The
term “inherited” is somewhat relative in the case of Romani as many of the words assumed to be
inherited with the described procedure will in fact be loanwords borrowed at an earlier stage in the
history of the language, and potentially present in (virtually) all dialects or branches. Romani has
borrowed extensively from other languages since its speakers' exodus from India (see for instance
Matras 2002), and even core parts of the lexicon are therefore borrowed. However, in cases where a
word is found to have cognates in different dialect branches, it is assumed to have been borrowed
into Romani such a long time ago that is has been fully adopted to Romani phonology, and is
therefore counted as an inherited word in this study. All of these observations are checked against
the etymological information in the mentioned literature. Some of the more difficult cases of this
procedure are described in section 4.1.
The second part of the process is then to analyse the inherited words in order to establish the
phonetics and phonology of Finno-Romani at the stage before it started adopting loanwords.
“Loanword” in this context refers only to words which are not found to have any plausible cognates
outside of Finno-Romani and which are not traceable as older loans. I do this by comparing Finno-
Romani with the two Sinti dialects found in the database: RO-022 and AT-001x. Using the
comparative method I determine the phonetic and phonological properties of the language stage
immediately before Sinti and Finno-Romani split into separate dialects (this stage will from now on
be referred to as 'Proto-Northwestern'). Proto-Northwestern is the closest it is possible to get to the
phonetics and phonology of Finno-Romani immediately before adoption of loanwords and
extensive contact with Swedish and Finnish. An easier solution would have been to simply use
Early Romani, which is the (already) reconstructed language stage of Romani after extensive
borrowing from Greek in somewhere around the tenth century (Elšik & Matras 2006: 68), but as I
wanted to get as close to the stage of Proto-Northwestern as possible, I deemed this too imprecise
18
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
and preferred to do my own reconstruction (I am also not aware of any actual word list with
reconstructed forms). This procedure is further described in section 5.
After having reconstructed the proto-stage and the sound changes which have happened
since then, I compare the changes revealed by the analysis with the sounds and sound structure of
Standard Spoken Finnish. Although the ideal subject of comparison would have been the Finnish
dialects spoken in the places where the various Finno-Romani dialects in the analysis are found, I
have chosen not to do so for several reasons. Firstly, very little material about the sounds and sound
structure is available on these Finnish dialects (especially in languages other than Finnish), which
would make for at comparison with great holes at best. Secondly, Standard Spoken Finnish is based
on all the Finnish dialects and is not - like many other standard varieties one particular dialect
which due to status has become the official language (Suomi, Toivanen & Ylitalo 2008: 7). Thirdly,
a very thorough description of the Finnish sound system is available (Suomi, Toivanen & Ylitalo
2008), and it seems preferable to have as basis a seemingly complete and thorough view of the
matter to a collection of partial sources which might not even agree. Fourthly, Suomi, Toivanen &
Ylitalo (2008) actually do comment on dialectic variations, and these comments will of course be
included whenever they occur and are deemed relevant. For these reasons I have decided that taking
Standard Spoken Finnish as basis for comparison is the most reasonable approach to the matter.
In many cases I use the forms of European Romani as listed by Matras (2010) for reference
in the analysis. I have “translated” these forms into broad IPA in line with the other transcriptions in
the paper, and they are listed in the final column in Appendix 2 for reference. Matras defines
“European Romani” as “the reconstructed entity referred to as 'Early Romani' (Elšik & Matras
2006; Matras 2002), along with any modifications that can be assumed to have emerged in the
particular forerunner dialect of British Romani. […] European loanwords that entered the language
prior to the isolation of British Romani and which are widely attested in other dialects of the
language are considered part of the European Romani legacy.” Although British Romani is an
independent dialect branch (Matras 2002: 10), its forerunner – European Romani – is considered
closely enough related to Finno-Romani to serve the present purpose. European Romani is a great
help in reconstructing Proto-Northwestern, as it provides information about a language stage which
comes before Proto-Northwestern and therefore makes it a lot easier to determine the direction of
sound changes. While it is possible to make reconstructions without knowledge of such a language
stage, it is more difficult and the results less certain. Given that the information is available, I
believe it would have been waste of work and knowledge not to use it and to build upon it. The
19
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
same goes for other published sources I use in the analysis: I see no reason not to make full use of
them, as long as this is done with a critical eye.
To my knowledge, no one has used exactly this procedure for a similar study. The
comparative method is obviously well-used, but as far as I can tell not with this goal. Though there
are aspects of this procedure which are perhaps not ideal, it is my claim that this is the best possible
approach to this topic.
4. Preliminary data analysis
In this section I give a detailed account of the first part of the analysis of Finno-Romani as
described in section 3.2. This account shows the process of determining which words in the RMS-
database' vocabulary lists of Finno-Romani are inherited and which are loanwords, and whatever
else is needed before starting the main analysis.
Appendix 1 shows the word lists for the Sinti dialects and Finno-Romani. Using existing
literature (mainly Boretzky 2012 and Valtonen 1972) and comparisons with the other dialects in the
RMS-database, all words in these lists have been sorted into three groups: words in multi-word
phrases (written in grey in Appendix 1 and not included in further analysis), non-cognates (written
in white on black and used sporadically for certain parts of the analysis and cognates (written in
black on white and used for the central part of the analysis (section 5)).
Several cases from the RMS-lists are not included in the table in Appendix 1. These are
cases where no word form from any dialect was useful: Cases with autonyms and cases with only
whole phrases instead of simple words. Compounds have also been excluded as these complicated
the initial analysis.
Many of the cases listed in Appendix 1 are further discussed in the following four sections:
Cases whose status as loanwords or inherited words is doubtful (section 4.1), cases where it seems
that there has been semantic shift (section 4.2), inflected words (section 4.4) and cases with more
than one cognate (section 4.3). These cases are all marked with a letter in the column labelled
'comments' in Appendix 1.
Cases where fewer than three dialects have inherited words (of the same cognacy) will not
be used in the analysis in section 5 but are nevertheless shown in Appendix 1 with the relevant word
forms noted as cognates.
The outcome of this preliminary data analysis – a list with cognate sets ready for analysis –
20
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
can be found in Appendix 2 and will work as the basis for the reconstruction described in section 5.
4.1. Cases which might be both loanwords and inherited words
I now briefly discuss the cases in which there is doubt about whether a word form is a loanword or
an inherited word. These cases are marked with the letter 'D' in the column 'Comments' in Appendix
1 and are discussed in numerical order here.
28 “cow” is a difficult case as the Finno-Romani forms [kurun], [kurni], [kurja], [gurunis]
look like both the inherited 'gurumni' but also German 'Kuh' and Swedish 'ko' which both mean
'Cow'. Boretzky (2012: 23) lists the form 'gurni' (which but for the voicing of the initial plosive is
identical with the forms of FIN-005 and FIN-008) as being a regular variation of the word, but only
in South-Balkan dialects. The suffixes -in and -ja are both feminine (see also section 4.4) and go
back to Sanskrit (-in, Valtonen 1972: 16). -ja is less certain, but probably a plural marker. It is
however clear that at least -ja has been productive even after contact with Germanic languages, so
the presence of old inflections does not rule out that these words might be loanwords. There is
nothing to explain how the 'r' present in all dialects would have been added to 'Kuh' or 'ko' though,
and for this reason, all the forms found in 28 “cow” have been counted as inherited words.
Case number 70 “belly” is quite clearly a Germanic loanword (could be from German
'Magen' or Swedish 'mage' for instance), but as the word is found in both Sinti and Finno-Romani, it
seems to have been borrowed before the two branches split, and it has therefore been included. the
same goes for 81 “sky”.
In 164 “lift” the forms [khantel] and [xandela] are found in FIN-002 and FIN-012
respectively. There is nothing to indicate that these forms are loanwords, and there is nothing in
their phonological forms that does not fit with Finno-Romani, but on the other hand none of the
other Romani dialects in the database have similar forms, and I have also not been able to find the
form described anywhere. For this reason they are marked as loanwords, but as there are only two
forms which might be useful, the case would not have been used in the analysis in any case.
In 185 “meet”, the forms [tikkel] and [tikka:] do not seem to be the standard Romani form,
but it turns out to be the word for “see”, which is case number 158. As the two are identical and 158
“see” is more complete, 185 “meet” is left out of the analysis.
In 186 “come”, RO-022 has the form [va] which at first glance does not look like it is
21
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
cognate with [a:vel] which is found in Finno-Romani. However, [avava] is very frequently found in
the other Romani dialects in the RMS-database, and this form can very easily have both [a:vel] and
[va] as its reflexes. I therefore keep both forms, but as Sinti is the dialects furthest from the most
commonly found form, its differences will not be discussed further.
Case number 197 “pay” is quite possibly not useful as a cognate. While there is a European
Romani form ples-, there is even at this point doubt whether or not this is influenced by the German
word for price, Preis (Matras 2010: 202). As this could have happened early enough to influence
European Romani but also easily at the stage of Proto-Northwestern or even later, it is impossible to
tell if the forms found in data reflect ples or preis and this case has therefore been excluded from the
analysis.
4.2. Cases with semantic shift
While it is outside the scope of this study to investigate semantic shifts in Romani, I will here
briefly mention the cases where a shift seems the most likely explanation and describe how each
case is handled in the analysis. Cases mentioned in this section are marked with the letter 'S' in the
column 'Comments' in Appendix 1.
In case number 46 “lip” FIN-002 has the form [mui] which is clearly an inherited word but
which did not originally mean “lip” but rather “mouth”. It still means “mouth” in case number 47 in
almost all the dialects, including FIN-002, which suggests that [mui] has undergone generalisation
in this dialect. Another possible explanation is that something in the field situation (such as the
researcher pointing) is responsible for this shift. As FIN-002 is the only dialect that uses [mui] to
mean “lip”, and as there are thus no other reflexes of the cognate that mean “lip”, the [mui] in 46
will simply not be included in the segmental analysis.
European Romani had two different words for “foot” (piro) and “leg” (heroj). Finno-Romani
uses forms of piro for both meanings (see cases 61 and 65). In 61 “leg” RO-022 also has a form of
piro, and the phrase RO-022 has in 65 “foot” also includes the word. This indicates that the same
generalisation has partially taken place in Sinti (AT-001x has the form [heri] in 65, so the
generalisation is clearly not complete in all the dialects), and it thus seems to be a generalisation
which has happened before or at the stage of Proto-Northwestern and therefore outside the scope of
this study. The two cases will be treated as two different cases as the different inflections of these
cases in FIN-006 suggests that it might not be two completely identical cases.
22
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
Matras (2010) lists no European Romani forms for “month” (84), but it is quite clear that all
the dialects in this study have generalised the word for “moon” (83) to cover the meaning of
“month” as well. As this does not seem to have happened after Proto-Northwestern, I will discuss it
no further, but the two cases have been kept separate as they show slight differences in form.
The word phuv meant “ground” in European Romani and not “earth” (which was the word
chik) although it is used in this meaning in all the dialects in this study (see 87). This is not a new
development and will not be treated further.
In 91 “rain” FIN-002 uses the form [pa:ni] which is also found in 90 “water in both FIN-002
and the other Romani dialects. Given that [pa:ni] has a much wider use as the form for “water” than
for “rain” in all the Romani dialects in the RMS-database, I assume that “water” is the original
meaning. I simply leave FIN-002 out of the analysis of 90 “water”.
The European Romani word mas meant “meat” while balavas meant “bacon”. While there is
no word for “meat” in the wordlists in the RMS-database, all the Finno-Romani dialects use [mas]
to mean “bacon” (case number 106). RO-022 has a completely different form in this case ([ʃuŋka])
which is from German Schinken, and it is possible that the change in meaning has happened only in
Finno-Romani. As the only potential conflict is in the meaning though, and as bacon is meat, 106
“bacon” is treated as any other case.
European Romani skamin meaning “chair” seems to be used in a generalised fashion in both
Finno-Romani and AT-001x to also cover the meaning “bench” (case number 126) in addition to
that of “chair” (case number 127). In both 126 and 127 RO-022 uses a different non-cognate which
is inflected differently in the two cases and for this reason I am not comfortable lumping the two
cases together even though the forms of skamin are identical.
Sinti uses the word [la:p] for both “word” and “name”. It is originally the word for “word”,
but its broader use is quite common among the Romani dialects (Boretzky 2012: 27). Sinti [la:p] is
not included in the analysis of 135 “name”.
In case 147 “light” Finno-Romani uses the form of the word which meant “fire” (see also
case 95) in European Romani. A generalisation of this word has thus taken place, but apparently
only in Finno-Romani. As 95 and 147 are not identical, they are treated separately.
Cases 180 “love” and 181 “want” both go back to the European Romani word kam- which
covered both meanings. It appears that the meaning of 182 “beg” has since been included as well, at
23
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
least in Finno-Romani. As none of the three cases are identical inflection-wise, they have all been
included in the analysis.
4.3. Cases with different cognates
In this section I discuss some of the cases whose word forms reflect more than one cognate. These
are all marked with the letter 'C' in Appendix 1. In cases where there are clearly two different
cognates and at least one of them have fewer than three reflexes across the dialects, this cognate
will not be used in the following analysis and will not figure in Appendix 2. For cases where this is
the only comment, there will be no further mention.
Case number 16 “horse” seems to have two quite different forms. Not only in Sinti and
Finno-Romani, but also across the other Romani dialects (approximations to either 'gras' or 'graj'). It
is of quite possible that the gras – graj difference has developed in Finno-Romani from a single
ancestor, but the presence of both forms in many other dialects makes it more likely that there are
two different cognates in game in Finno-Romani. As only FIN-008 has the form ending with a
fricative, this occurrence will simply be taken out of the analysis and the other form will be
analysed in the regular way.
Already at the stage of European Romani the word “arm” had two different forms: musi and
vast (Matras 2010: 189) with vast being the most general term meaning both “hand” and “arm” and
musi being more specialised and meaning only “arm”. Various forms of vast still mean “hand” in all
the dialects of this study, but FIN-005, FIN-008 and AT-001x use forms of musi for “arm” instead.
Given that this division is found already in European Romani, it seems clear that this specialisation
happened before the Proto-Northwestern stage and is therefore not the concern of this study. As
there are three instances of both [vast] and [mus-] in 55, they are simply treated separately so that
the forms of [vast] are called 55a and the forms of [mus-] 55b in Appendix 2.
European Romani had a word for “tree” (rukh) and one for “wood” (kašt), but while all the
dialects in the study still use forms of kašt to mean “wood”, FIN-005, FIN-006 and FIN-012 also
use this to mean “tree” while the Sinti dialects and FIN-008 have kept forms of rukh. As there are
three of each form in 96, this case has simply been split up into two separate cases: 96a and 96b.
There are certain similarities in 146 “gold” between Sinti [symnokaj] and Finno-Romani
[sonatiko] (and its variations), but they also look similar enough to possibly be cognates. European
Romani has the form [sonakaj] which appears to be a mixture of the two forms. Both forms, and
24
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
also forms that look like mixtures of the two, are found in the other Romani dialects in the RMS-
database. It seems then that the two forms are cognates, but that the differences between the two go
back much further than Proto-Northwestern, and I therefore treat them as different cognates.
The same is true for 190 “leave sth”, in which [mukk-] and [mekk-] appear to reflect an
older split (Boretzky 2012), for [arre] and [an] in 216 “in”, and for the Sinti [dis] in 77 “day”.
In case number 199 “bad (nasul)” RO-022 [dʒungalo] appears similar enough to the Finno-
Romani dialects ([pengalo]) that it seems fair to suggest similarity of form. However, as it turns out,
beng is the word for “devil” in European Romani while džungalo originally meant “ugly”. The only
similar thing is -alo, which is an adjectivising suffix. With those two cognates thus separated, there
are not enough occurrences of the form to be included in the analysis.
4.4. Morphology
In this section I make a few observations of inflectional and derivational morphology found in the
inherited words in the data set. The aim is not to analyse the history of morphology in Finno-
Romani in itself, and it is outside the scope of this study to attempt an explanation as to why which
suffixes go on which words in which dialects. Instead, the goal is to use whatever morphological
clues are available to further the phonological analysis, and most of all to avoid drawing
conclusions about sound changes which are really a matter of different suffixes. All suffixes have
therefore been removed both in cases where they were identical and in cases where they were
different across the dialects. The morphemes themselves are not discussed here, but such discussion
can be found by for instance Matras (2002).
4.5. Other considerations
In a few cases there are circumstances to note which are not already presented.
In 59 “finger” there are very clearly reflexes of the same cognate in all the dialects.
However, these forms are unusually different compared to the rest of the cognates, which makes it
almost impossible to know which segments should be matched up across the dialects. In order to
keep the analysis as straight-forward as possible, I have only included the segments where there was
no doubt in Appendix 3.
In 92 “snow” the original RO-022 [ijp] from the RMS-database look like a transcription
mistake. Judging by the sound file, there is nothing wrong with the transcription, but I find that [i:p]
25
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
is a more reasonable interpretation than the original [ijp]. This makes the case much more
straightforward to handle in the analysis.
5. Data analysis
In this section I give a detailed account of the second part of the analysis of Finno-Romani as
described in section 3.2, namely the analysis of the inherited words in Finno-Romani. The goal is to
describe the language stage before Sinti and Finno-Romani split into different dialects (Proto-
Northwestern), and the sound changes which have happened since then. I present the segmental
inventory of Proto-Northwestern and the sound changes that have occurred since that stage
(sections 5.1 and 5.3. In order to provide an idea of the context these segments are found in, I also
briefly discuss stress and length (section 5.5).
I first reconstruct and discuss the segmental inventory of Proto-Northwestern, which can be
found in sections 5.1 and 5.3. To do so, I follow the procedure for using the comparative method as
described by Crowley & Bowern (2010), starting with the setting up of correspondence sets for each
cognate. An overview of all the correspondences can be found in Appendix 3. This is complicated
by the many cases where one or more dialects for various reasons have gaps and therefore no
correspondence. This sometimes makes it impossible to determine which type of correspondence a
cognate set shows. Appendix 3 therefore simply shows all the correspondence sets with gaps, but
the sets are grouped together according to the groups they most likely belong too (For instance, the
set a;a;a;-;a;a;a most likely belongs to the set a;a;a;a;a;a;a as there are no obvious alternatives). In
cases where there is more than one possibility, a set is grouped with other sets according to what
seems most plausible, and this will be discussed in the following. Each group of correspondence
sets in Appendix 3 is named for the protophoneme which I judge it to be reflexes of (left column).
In cases where a protophoneme has sets of reflexes which are clearly different, these are grouped
separately. All groups are also given a number in addition to the protophoneme for reference. One
group of correspondences is marked with “N” as the relevant segments are different kinds of nasals
– the N symbolizes an unspecified nasal consonant. Length is marked as a property of the vowels
(using the standard IPA sign ” : ”) and long vowels are sorted together in the correspondence sets.
This is done because the correspondence sets will serve as basis for the analysis and reconstruction
of both the segmental inventory as well as stress and length. Terminology relating to language
change are used as per Crowley & Bowern 2010.
26
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
For each part of the analysis (vowel inventories, consonant inventories, length and stress), I
compare the changes found with the sound structure of Standard Spoken Finnish (see 3) in an
attempt to determine if or to which degree the change is motivated by contact with Finnish and also
to discover which possible contact-induced changes have not taken place.
It is outside the scope of this study to investigate phonotactics and syllable structure, but I
use the following guideline in cases where it is useful to know the position of syllable boundaries:
“every CV sequence is preceded by a syllable boundary” (Granqvist 2002: 74).
5.1. Analysis of correspondences with vowels
In this section I present my reconstruction of the vowels of Proto-Northwestern and the changes
which they have undergone since then. Firstly (this section), all the correspondences are presented
briefly along with my analysis of them, and secondly (section 5.2), the results of the analysis are
discussed and compared with Standard Spoken Finnish. Matters connected to length and stress as
well as correspondences marked 'Diphthongs' are discussed separately in section 5.5.
[a1-2+5-7] First of all, many words contain the sound [a] across all dialects. These appear to be
unchanged since at least Proto-Northwestern and in many cases also since European Romani as they
are also found there. This correspondence can be assumed to reflect *a.
[a3+8] In three cases (186 “come”, 215 “outside” and 220 “we”), RO-022 has nothing, i.e zero (for
instance in 215 [vri]) where the Finno-Romani dialects have [a] (for instance FIN-012 [auri]). This
occurs in word-initial position in both cases. In 220 “we”, FIN-012 too does not have an [a] ([me])
while FIN-002, FIN-005 and FIN-008 do ([ame]).
While RO-022 only rarely has [a] in word-initial position, it does occur (even among
inherited words) and there is no clear system as to when an [a] disappears word-initially and when it
does not. It is clear, however, from comparison with the other Romani dialects and European
Romani that it is a case of apheresis (loss of an initial segment) in RO-022 and not of prothesis
(insertion of an initial segment) in Finno-Romani. FIN-012 has word-initial [a]'s on several
occasions, and there is no clear reason why it has been lost in 220 “we”. The correspondences of
this group then also reflect *a. Although there are no clear conditions for the rule(s) that account for
the change in RO-022 and FIN-012, the following rule does seem to have taken some effect in both
27
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
dialects:
*a → Ø / #_ (FIN-012 (Partial: 220) and RO-022 (Partial: 186, 215, 220))
[a4] In two cases (11 “girl” and 16 “horse”), the sound [æ] appears in some of the Finnish Romani-
dialects where the rest of the dialects have [a] (for example FIN-005 [ʦæi] versus FIN-002 [ʦai] in
11). This looks like an example of assimilation (that one vowel has adopted one or more features –
in this case it has become higher - from a neighbouring vowel) as [a] in these cases are next to [i],
but as there are only two cases which are not even completely alike, it is not possible to set up a
general rule for vowel assimilation in Finno-Romani. It does however seem clear that it can be
found in FIN-005 and FIN-012. Although 82 (“star”) appears to be an exception as FIN-005 has the
form [ʦerxai] and not [ʦerxæi], this can be explained by the fact that the diphthong [ai] in 82
(“star”) appears in an unstressed syllable. If the assimilation appears only in stressed syllables, there
are no counter examples. Although FIN-008 only has a single case (11 “girl”) of vowel harmony, it
too seems to have undergone a similar sound change, although it is found next to [j] rather than [i]
in this case. Given how closely [i] and [j] resemble each other articulation-wise, it seems reasonable
to combine them into the same rule.
FIN-006 too seems to have undergone this change judging by 11(“girl”), but 16 (“horse”) is
a clear counter example ([trai]). As there is no logical reason for 16 (“horse”) to have [a] and not
[æ] (or indeed for 11 (“girl”) to have [æ] rather than [a] given that there are only the two cases
where the rule might potentially apply), it is not possible to determine the exact course of change in
FIN-006. It does however seem clear that these correspondences reflect *a as well. The final rule
then looks like the following:
*a [+stress] → æ / _ i,j (FIN-005 (Complete), FIN-006 (Partial: 11) FIN-008 (Partial: 11),
FIN-012 (Partial: 11, 16))
[u1+3] When it comes to correspondences with [u], there are a few cases with plain [u]'s across all
the dialects. There are not nearly as many of these as there are cases with plain [a] across all the
dialects, but it still seems reasonable to reconstruct a *u for Proto-Northwestern for these cases and
not note any changes.
28
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
[v1] Another group consists of cases which have [u] in the Finno-Romani dialects and [p] in RO-
022 - for instance in case number 132 (“word”) which is [lau] in FIN-012 and [la:p] in RO-022. A
comparison with all the other dialects in the database shows that the most widespread form of the
[u]/[p] segment is [v] (which is also found in European Romani), and it seems plausible that this
was also the case in Proto-Northwestern. Since then *v must then have undergone rounding and
become syllabic in Finno-Romani ([u]) and have hardened to a plosive in RO-022 ([p]). This only
happens when the original *v occurs in syllable-final position as shown by 186 “come” where FIN-
006 has [aulo] while FIN-005 has [a:vel]. The syllable boundary is assumed to be between [u] and
[l] in [au.lo] and between [a:] and [v] in [a:.vel]. While this mostly happens after [a], 92 (“snow”)
shows that it can also happen after [i], and 87 “earth” shows that it can also happen after [u] in
which case [u] simply becomes long in Finno-Romani. This then seems to be a general rule that
states that any *v in syllable-coda becomes [u] in Finno-Romani. The result is two new diphthongs,
[au] and [iu], and lengthening of a previously short vowel, [u:], in all the Finno-Romani dialects
with the partial exception of FIN-005 (where case number 132 “word” has to forms: [la:v] and
[lau]).
Case number 215 “outside” suggests that the original *v has stayed a [v] in RO-022 when it
is not word-final (the RO-022 form is [vri]), but as there is only one case where this segment is not
word-final and as Sinti is included in the analysis mainly in order to make the analysis of Finno-
Romani clearer, this will not be discussed further. For this group I then reconstruct a *v.
Whether the conditioning factor for *v to become [u] is the fact that *v occcurs in syllable-
coda or if it is simply the fact that it occurs immediately after a vowel is not entirely sure. If the
condition is only that *v occurs after a vowel, then this rule would predict that forms like 152
[ji:vela] or 186 [a:vel] would instead be [ji:uela] and [a:uel] respectively. This is obviously not the
case. However, in both these two cases [v] is found after long vowels, therefore it is more likely that
they stand in syllable-onset. This again points back to the condition which is concerned with *v's
position in the syllable rather than its direct contact with a vowel. There are no cases where *v
occurs in syllable-coda but not immediately after the vowel and indeed this might not be permitted
by the phonotactics of Finno-Romani, in which case it is a somewhat moot point whether or not the
rule ought to allow for this. Given the available data I prefer a rule that states that *v changes to [u]
when it occurs in syllable-coda for the reason that it explains all the documented occurrences.
Whether or not its claim is too broad must be reconsidered if any cases show up with the structure
VCv in a single syllable. Until this happens, I note the following rule for Finno-Romani:
29
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
*v → u / _ . (Complete with one partial exception: FIN-005: 132)
[u2] Case number 10 “boy” appears to show a situation where either an original *u has become [o]
in RO-022 making the the Sinti form [tʃao] (RO-022), or an original *o has become [u] in Finno-
Romani making the Finno-Romani form [tʃau] (FIN-012). While this is not an impossible sound
change, a comparison with all the other Romani dialects in the RMS-database and the European
Romani forms strongly suggests that the original form of the word was actually [ʧavo]. This makes
another course of events quite likely: First [o] disappeared in Finno-Romani (-o is a masculine
suffix) and then the resulting [ʧav] underwent the rule stated above where a syllable-final [v]
becomes [u]. RO-022 appears to have lost [v] instead, but I will not discuss this further. The [u]'s in
this correspondence group then reflect *v rather than *u or *o.
[e1+2] As for correspondences with [e], they seem to be quite stable generally, as there are no
changes at the segment level across the dialects in the e-correspondences. I therefore reconstruct *e
for both these correspondence groups without further discussion.
[o1+2] This stability is also found in many cases with o-correspondences: [o] is found across all the
dialects, and it seems reasonable to reconstruct *o for these cases. In the RMS-database case
number 234 (“white”) is transcribed as having [ø] rather than [o] in FIN-012, but I have judged this
to be a transcription error and have noted it as [o] instead. I have based this decision on the fact that
232 (“short”) too is noted as having [ø] in FIN-012, and in that word ([støt]) it is very clearly the
case while in 234 (“white”) I for one cannot hear it. I thus do not note any changes for these groups.
[o3] In two cases, 83 (“moon”) and 84 (“month”) FIN-002 has [u] where the other Finno-Romani
dialects and Sinti have [o] (for instance FIN-002 [tʃu:n] and FIN-012 [tʃho:n]). A comparison with
the other dialects in the RMS-database shows that these two cases are quite clearly reflexes of the
same cognate even though they have different meanings. The two words also show up with identical
forms in FIN-008, FIN-012 and AT-001x (and nearly identical forms in FIN-005, while FIN-006 has
a gap and might also potentially have had an identical form). This is confirmed by Boretzky (2012:
42). Almost all other dialects where this form is found in either 83 (“moon”) or 84 (“month”) have
30
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
[o] rather than [u]. The most obvious explanation of these facts is that [u] and [o] in 83 (“moon”)
and 84 (“month”) are both to be regarded as reflecting *o and that something has possibly happened
in FIN-002 to change this. As it is not possible to verify the transcription and as there are only these
two cases which more or less only count for one, I will not attempt to explain this any further, and
do not set up a rule for it.
[i1+2] Many cases show [i] across all the dialects, and it seems obvious to reconstruct *i for these
cases.
[j1] Three cases (11 “girl”, 16 “horse”, 47 “mouth”) show a show a pattern where FIN-002, FIN-
005, FIN-006 and FIN-012 have [i] while FIN-008 and the two Sinti dialects have [j] – for instance
is 47 [mui] in FIN-012 and [muj] in RO-022. In all three cases [i]/[j] appears right after a vowel and
thus becomes the last part of a diphthong. A comparison with European Romani and all the other
Romani dialects shows that the segment was originally [j] and these three cases then show a
development parallel to [v] becoming [u] as shown earlier in this section: Certain consonants
become parts of diphthongs in Finno-Romani when they appear in syllable coda. However, unlike
the v → u change, this does not appear to happen in FIN-008 for some reason, although as there are
only two cases to show that it does not and [i] and [j] can be hard to tell apart in syllable-codas, this
might in fact be a general rule for Finno-Romani. Another explanation might be that the various
transcribers of the dialects in the RMS-database have used slightly different transcription
conventions. This does not seem particularly likely given the amount of work which has gone into
unifying these conventions though (see Matras, White & Elšik 2009: 11), so I am simply noting that
FIN-008 appears to behave differently in this regard:
*j → i / _ . (Complete except in FIN-008)
There are several correspondence sets where a vowel segment appears in one dialect but not in
others. In order to decide whether these groups show sound addition or sound loss, I compare them
to the forms listed by Boretzky and to the European Romani forms listed by Matras (2010). In cases
where there was no European Romani form listed, I have compared the Finno-Romani forms with
the other Romani dialects in the RMS-database and noted the most widespread form as an
31
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
approximation of a protophoneme.
[Ø1] In two cases a dialect has added [e] word-finally. In 6 “friend” it is FIN-006 ([ma:le]) and in
84 (“month”) it is FIN-005 ([ʦo:ne]). As it does not appear to be a general tendency in either of the
dialects and as it happens in two different dialects but in both cases to nouns, it seems likely that
this might be a morphological matter. There are several Romani nominal suffixes with the form -e
but these are either plural markers or vocative suffixes (Matras 2002: 82-85) and it is not clear why
the words in the given cases would display either outside of any context. However, as there are also
no apparent reason for the variation in the segmental structure, it seems more likely that it should be
a morphological matter and I will therefore not treat it any further.
[Ø2] In 206 “now” Finno-Romani has lost a word-final [a] (also a word-initial one, but that
development is shared by Sinti and is thus assumed to have happened before Northwestern split into
a separate branch and is therefore not relevant here) which Sinti has kept ([ka:na]). As this is the
only case where this has happened and as the word can be assumed to be quite frequent and
therefore vulnerable to sound loss, I will not deal any further with this than to note that the
Northwestern protoform of the word most likely has an *a.
[Ø3] The European Romani form of 139 “pocket” reveals that the original form is kept in Sinti (The
form [posita] which is also listed as common in the Northern dialects by Boretzky (2012: 71), while
it has lost a syllable in Finno-Romani, giving it the form [posta]. 139 “pocket” is unusual in that the
stress is on the second syllable instead of the first in RO-022, and this might have been the reason
that this word alone has lost a word-medial [i]. The most likely course of change is that the stress
moved to the first syllable and that the [i] was consequentially lost (syncope), but as there is only
this one case, it is not possible to say for certain. It seems reasonable to construct *i for this
segment, but as it appears to be a unique example, I will not attempt to state the rule governing this
change.
[Ø4] In 85 “wind” FIN-012 has acquired an extra [a] word-medially ([balava]). As with 139
“pocket” this seems to be a unique case, and I will not state any rules for it. Judging by European
Romani and the other dialects in the RMS-database, this word did not have *a in the middle of the
protoform, and according Granqvist (in press) mentions the case as an example of a quite recent
32
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
process of epenthesis of a svarabakhti vowel in certain words in Finno-Romani.
5.2. Discussion of vowel inventories
In summary of the analysis of correspondences with vowels, it seems, then, that on a segmental
level, no great changes have occurred in the vowels of Finno-Romani since Proto-Northwestern. No
segments have been lost and no new segments have been added. Apart from [v] becoming [u] and
[j] becoming [i] post-vocally in Finno-Romani (which are not vowel changes in a strict sense) there
are only minor, occasional differences between RO-022, AT-001x and the Finno-Romani dialects to
be found, such as the vowel harmony happening in FIN-005 and FIN-012 and to some extent in
FIN-008.
Proto-Northwestern can then be assumed to have had a standard symmetrical five-vowel
system with three heights:
Figure 1 – The vowel system of Proto-Northwestern
While the overall tendency in the cognates seems to be that the vowels have been quite stable in the
Northwestern group of Romani, the picture is a little different when it comes to simple comparison
between the various modern dialects and Proto-Northwestern. All the modern dialects still have all
the vowel segments shown above, but when considering all the inherited words listed in Appendix 1
and not just the ones in Appendix 2, it becomes clear that all the dialects except FIN-002 also have
[æ]. When including the borrowed words as well, FIN-002, FIN-006 and FIN-012 also have [y] and
[ø], and FIN-002 also has [æ] although these are never found in the inherited words. FIN-005 lacks
[y] but does have [ø], whereas FIN-008 seems to be closest to the vowel inventory of Proto-
Northwestern as it has only added [æ]. Generally, then, although there are variations amongst the
33
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
dialects, the modern dialects of Finno-Romani have gained three front vowels - [y], [ø] and [æ] - in
addition to the original five vowels of Proto-Northwestern. This makes for a quite atypical vowel
system which does not appear to be symmetrical:
Figure 2 – Vowel inventory of most of the modern Finno-Romani dialects
Although there have been relatively few actual changes to the vowels in Finno-Romani, the modern
dialects thus do have vowel inventories which are quite different from that of Proto-Northwestern,
although not all the segments in the inventories are found in the inherited material. The vowel
inventory shown above in figure 2 appears to be identical to that of Standard Spoken Finnish
(Suomi, Toivanen & Ylitalo 2008: 21). Of this inventory Suomi, Toivanen & Ylitalo (2008) note the
following: That the mid-series /e/, /ø/ and /o/ are actually halfway between IPA [e] and [ɛ], [ø] and
[œ] and [o] and [ɔ] respectively, that /a/ is about the same height as /æ/, and that the inventory is
more symmetrical than it immediately appears: There are three height classes (/i/, /y/ and /u/; /e/, /ø/
and /o/; and /a/ and /æ/) and “no vowels not assignable phonetically to a class consisting of at least
two vowels“ (Suomi, Toivanen & Ylitalo 2008: 22). Furthermore, the Finnish vowel system is
described as being very stable and with barely any variation across dialects and registers (Suomi,
Toivanen & Ylitalo 2008: 23). It would require a more specifically phonetic study and a finer
transcription to verify if all these traits are shared by Finno-Romani, but at the phonological level,
34
Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012.
these two languages do share the same vowel inventory.
That is not to say that this is necessarily due to Finno-Romani being in contact with Finnish.
(Valtonen 1968, citet via Granqvist 2002) suggests that contact with Hungarian has given Finno-
Romani [y] and [ø], and that [æ] is a later loan from the Scandinavian languages. Granqvist does
not comment on this (except to note that the sequence [æi] resembles old Vlax-dialects spoken in
Russia, but he does not seem to be suggesting that there is any connection between the two
(Granqvist 2002: 64)), but given the fact that [æ] is more widely found in the inherited lexicon than
both [y] and [ø], it does not seem immediately obvious that [æ] should be the newest loan. This
does not, of course, rule out the possibility that Valtonen is right at this point. It is also possible that
[æ] simply is more easily incorporated in words with the phonology of Proto-Northwestern.
As for [y] and [ø] coming from the Scandinavian languages (which I understand to mean the
Germanic languages of Denmark and the actual Scandinavian peninsula), this is perfectly possible
as they all do have the sounds in question or at least some that resemble them closely. There are also
loanwords in Finno-Romani which are clearly of Scandinavian origin which do have some of these
sounds (for instance 124 “bed” which in FIN-006 is [sæŋøs] although the same word has [e] and [o]
instead of [æ] and [ø] in both FIN-002 and FIN-012), although I cannot find any Scandinavian
words in the list containing [y].
While there thus might have been ample influence from the Scandinavian languages, I find it
remarkable that the vowel inventories of Finnish and Finno-Romani resemble each other as closely
as they do. Even is the [y], [æ] and [ø] originally entered Finno-Romani via the Scandinavian
languages or Hungarian, the similarities suggest that Finnish has played and still plays a significant
role in the shaping of Finno-Romani. Even if Finnish is not the source of the enrichment of the
vowel inventory of Finno-Romani, it might be what prevents that the new segments are lost again.
All of that being said, it is also clear that these changes involving the new vowels have only
taken partial effect, as the inventory of the inherited material has only been affected to a small
extent (as opposed to the rule turning syllable-final [v]'s to [u]'s which is found everywhere). Also,
as was noted above, figure 2 is only true for the collected Finno-Romani dialects – not for each of
the dialects on its own. I conclude, then, that having a given vowel appear in loanwords does not
necessarily mean that it has also been adopted to be used in inherited words. This also explains why
the reconstruction did not reveal any major changes: The sounds in borrowed words apparently do
not automatically or directly spread to the inherited lexicon. In the case of [æ] though, it seems that
a borrowed sound has spread. This might be a matter of it having been present in Finno-Romani for
35
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology
Contact Effects in Phonology

More Related Content

What's hot

Linguistics firse term exam
Linguistics firse term examLinguistics firse term exam
Linguistics firse term examcarelu
 
Introduction to linguistics lec 1
Introduction to linguistics lec 1Introduction to linguistics lec 1
Introduction to linguistics lec 1Hina Honey
 
Linguistics Theories MPB 2014 Progressive-edu.com
Linguistics Theories MPB 2014  Progressive-edu.comLinguistics Theories MPB 2014  Progressive-edu.com
Linguistics Theories MPB 2014 Progressive-edu.comHono Joe
 
Ujian compre
Ujian compreUjian compre
Ujian comprebharoes
 
Text Linguistics
Text LinguisticsText Linguistics
Text Linguisticsguest25b10e
 
Discourse Analysis Presented To Miss Rabia
Discourse Analysis Presented To Miss RabiaDiscourse Analysis Presented To Miss Rabia
Discourse Analysis Presented To Miss RabiaDr. Cupid Lucid
 
Assimilation and reduplication in pangasinan adjectives
Assimilation and reduplication in pangasinan adjectivesAssimilation and reduplication in pangasinan adjectives
Assimilation and reduplication in pangasinan adjectivesshinathrun
 
Discourse analysis.its development and application
Discourse analysis.its development and applicationDiscourse analysis.its development and application
Discourse analysis.its development and applicationAbdullah Saleem
 

What's hot (12)

Linguistics firse term exam
Linguistics firse term examLinguistics firse term exam
Linguistics firse term exam
 
Introduction to linguistics lec 1
Introduction to linguistics lec 1Introduction to linguistics lec 1
Introduction to linguistics lec 1
 
Linguistics Theories MPB 2014 Progressive-edu.com
Linguistics Theories MPB 2014  Progressive-edu.comLinguistics Theories MPB 2014  Progressive-edu.com
Linguistics Theories MPB 2014 Progressive-edu.com
 
C371321
C371321C371321
C371321
 
Ujian compre
Ujian compreUjian compre
Ujian compre
 
Discourse Analysis
Discourse AnalysisDiscourse Analysis
Discourse Analysis
 
Text Linguistics
Text LinguisticsText Linguistics
Text Linguistics
 
Introduction to linguistics
Introduction to linguisticsIntroduction to linguistics
Introduction to linguistics
 
Discourse Analysis Presented To Miss Rabia
Discourse Analysis Presented To Miss RabiaDiscourse Analysis Presented To Miss Rabia
Discourse Analysis Presented To Miss Rabia
 
Hstorical Linguistics
Hstorical LinguisticsHstorical Linguistics
Hstorical Linguistics
 
Assimilation and reduplication in pangasinan adjectives
Assimilation and reduplication in pangasinan adjectivesAssimilation and reduplication in pangasinan adjectives
Assimilation and reduplication in pangasinan adjectives
 
Discourse analysis.its development and application
Discourse analysis.its development and applicationDiscourse analysis.its development and application
Discourse analysis.its development and application
 

Viewers also liked (20)

Impedimenta
ImpedimentaImpedimenta
Impedimenta
 
Natatorium state emails
Natatorium state emailsNatatorium state emails
Natatorium state emails
 
Triskaidekaphobia
TriskaidekaphobiaTriskaidekaphobia
Triskaidekaphobia
 
Kenkoh Massage Sandals
Kenkoh Massage SandalsKenkoh Massage Sandals
Kenkoh Massage Sandals
 
Z barriers
Z barriersZ barriers
Z barriers
 
Patrician
PatricianPatrician
Patrician
 
Oogénesis
OogénesisOogénesis
Oogénesis
 
Intrapleural stk
Intrapleural stkIntrapleural stk
Intrapleural stk
 
Materializing the Web of Linked Data
Materializing the Web of Linked DataMaterializing the Web of Linked Data
Materializing the Web of Linked Data
 
Selenologytoday24
Selenologytoday24Selenologytoday24
Selenologytoday24
 
Gramatica inglesa
Gramatica inglesaGramatica inglesa
Gramatica inglesa
 
C&c final project (people & space)
C&c final project (people & space)C&c final project (people & space)
C&c final project (people & space)
 
Dr Hulusi Behçet - Turkish Dermatologist
Dr Hulusi Behçet - Turkish DermatologistDr Hulusi Behçet - Turkish Dermatologist
Dr Hulusi Behçet - Turkish Dermatologist
 
Security and information assurance
Security and information assuranceSecurity and information assurance
Security and information assurance
 
Cycle biologique de Plasmodium: du gamétocyte au zygote
Cycle biologique de Plasmodium: du gamétocyte au zygoteCycle biologique de Plasmodium: du gamétocyte au zygote
Cycle biologique de Plasmodium: du gamétocyte au zygote
 
Artificial Insemination in Alpacas
Artificial Insemination in AlpacasArtificial Insemination in Alpacas
Artificial Insemination in Alpacas
 
Fiac
FiacFiac
Fiac
 
First gastrectomy 1964
First gastrectomy 1964First gastrectomy 1964
First gastrectomy 1964
 
Nutritional anemias
Nutritional anemiasNutritional anemias
Nutritional anemias
 
Clansman Dynamics
Clansman DynamicsClansman Dynamics
Clansman Dynamics
 

Similar to Contact Effects in Phonology

Seminar applied linguistics
Seminar applied linguisticsSeminar applied linguistics
Seminar applied linguisticsHani Shakir
 
Applied linguistic: Contrastive Analysis
Applied linguistic: Contrastive AnalysisApplied linguistic: Contrastive Analysis
Applied linguistic: Contrastive AnalysisIntan Meldy
 
Branches of Linguistics Presentation Detail
Branches of Linguistics Presentation DetailBranches of Linguistics Presentation Detail
Branches of Linguistics Presentation DetailNomanButt30
 
The relevance of phonology to communication studies
The relevance of phonology to communication studiesThe relevance of phonology to communication studies
The relevance of phonology to communication studiesAlexander Decker
 
Dialnet finnishenglish phoneticsandphonology-234248
Dialnet finnishenglish phoneticsandphonology-234248Dialnet finnishenglish phoneticsandphonology-234248
Dialnet finnishenglish phoneticsandphonology-234248huu_trinh
 
Learning about language structure
Learning about language structureLearning about language structure
Learning about language structureRoda Menil
 
The Input Learner Learners Forward Throughout...
The Input Learner Learners Forward Throughout...The Input Learner Learners Forward Throughout...
The Input Learner Learners Forward Throughout...Tiffany Sandoval
 
Macro Sociolinguistics Insight Language (1).pdf
Macro Sociolinguistics Insight Language (1).pdfMacro Sociolinguistics Insight Language (1).pdf
Macro Sociolinguistics Insight Language (1).pdfFreddyBenjaminSepulv
 
Introduction to linguistics lec 1
Introduction to linguistics lec 1Introduction to linguistics lec 1
Introduction to linguistics lec 1Hina Honey
 
. . . all human languages do share the same structure. More e.docx
. . . all human languages do share the same structure. More e.docx. . . all human languages do share the same structure. More e.docx
. . . all human languages do share the same structure. More e.docxadkinspaige22
 
. . . all human languages do share the same structure. More e.docx
 . . . all human languages do share the same structure. More e.docx . . . all human languages do share the same structure. More e.docx
. . . all human languages do share the same structure. More e.docxShiraPrater50
 
Definitions, Origins and approaches of Sociolinguistics
Definitions, Origins and approaches of Sociolinguistics Definitions, Origins and approaches of Sociolinguistics
Definitions, Origins and approaches of Sociolinguistics AleeenaFarooq
 
General linguistics
General linguisticsGeneral linguistics
General linguisticszhian asaad
 
Derivational process in matbat language (jurnal ijhan)
Derivational process in matbat language (jurnal ijhan)Derivational process in matbat language (jurnal ijhan)
Derivational process in matbat language (jurnal ijhan)Trijan Faam
 
Makalah Phonological Construction
Makalah Phonological ConstructionMakalah Phonological Construction
Makalah Phonological ConstructionJerusman Marbun
 

Similar to Contact Effects in Phonology (20)

Seminar applied linguistics
Seminar applied linguisticsSeminar applied linguistics
Seminar applied linguistics
 
Applied linguistic: Contrastive Analysis
Applied linguistic: Contrastive AnalysisApplied linguistic: Contrastive Analysis
Applied linguistic: Contrastive Analysis
 
Branches of Linguistics Presentation Detail
Branches of Linguistics Presentation DetailBranches of Linguistics Presentation Detail
Branches of Linguistics Presentation Detail
 
The relevance of phonology to communication studies
The relevance of phonology to communication studiesThe relevance of phonology to communication studies
The relevance of phonology to communication studies
 
Dialnet finnishenglish phoneticsandphonology-234248
Dialnet finnishenglish phoneticsandphonology-234248Dialnet finnishenglish phoneticsandphonology-234248
Dialnet finnishenglish phoneticsandphonology-234248
 
Learning about language structure
Learning about language structureLearning about language structure
Learning about language structure
 
Language descriptions
Language descriptionsLanguage descriptions
Language descriptions
 
The Input Learner Learners Forward Throughout...
The Input Learner Learners Forward Throughout...The Input Learner Learners Forward Throughout...
The Input Learner Learners Forward Throughout...
 
Macro Sociolinguistics Insight Language (1).pdf
Macro Sociolinguistics Insight Language (1).pdfMacro Sociolinguistics Insight Language (1).pdf
Macro Sociolinguistics Insight Language (1).pdf
 
Introduction to linguistics lec 1
Introduction to linguistics lec 1Introduction to linguistics lec 1
Introduction to linguistics lec 1
 
. . . all human languages do share the same structure. More e.docx
. . . all human languages do share the same structure. More e.docx. . . all human languages do share the same structure. More e.docx
. . . all human languages do share the same structure. More e.docx
 
. . . all human languages do share the same structure. More e.docx
 . . . all human languages do share the same structure. More e.docx . . . all human languages do share the same structure. More e.docx
. . . all human languages do share the same structure. More e.docx
 
Definitions, Origins and approaches of Sociolinguistics
Definitions, Origins and approaches of Sociolinguistics Definitions, Origins and approaches of Sociolinguistics
Definitions, Origins and approaches of Sociolinguistics
 
General linguistics
General linguisticsGeneral linguistics
General linguistics
 
Derivational process in matbat language (jurnal ijhan)
Derivational process in matbat language (jurnal ijhan)Derivational process in matbat language (jurnal ijhan)
Derivational process in matbat language (jurnal ijhan)
 
Makalah Phonological Construction
Makalah Phonological ConstructionMakalah Phonological Construction
Makalah Phonological Construction
 
Hstorical Linguistics
Hstorical LinguisticsHstorical Linguistics
Hstorical Linguistics
 
Hstorical Linguistics
Hstorical LinguisticsHstorical Linguistics
Hstorical Linguistics
 
The study of language
The study of languageThe study of language
The study of language
 
Linguistics (Report)
Linguistics (Report)Linguistics (Report)
Linguistics (Report)
 

Contact Effects in Phonology

  • 1. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. Contact Effects in Phonology: A Case-study of Finno-Romani Index Abstract ............................................................................................................................................ 2 1. Introduction ................................................................................................................................. 2 2. Background .................................................................................................................................. 5 2.1. Sociolinguistic explanations ................................................................................................ 5 2.2. Structural explanations ........................................................................................................ 9 2.3. Typological explanations ................................................................................................... 12 2.4. Summary ............................................................................................................................ 13 3. Method ....................................................................................................................................... 14 3.1. Data .................................................................................................................................... 15 3.2. Procedure ........................................................................................................................... 17 4. Preliminary data analysis ........................................................................................................... 20 4.1. Cases which might be both loanwords and inherited words .............................................. 21 4.2. Cases with semantic shift ................................................................................................... 22 4.3. Cases with different cognates ............................................................................................ 24 4.4. Morphology ....................................................................................................................... 25 4.5. Other considerations .......................................................................................................... 25 5. Data analysis .............................................................................................................................. 26 5.1. Analysis of correspondences with vowels ......................................................................... 27 5.2. Discussion of vowel inventories ........................................................................................ 33 5.3. Analysis of correspondences with consonants ................................................................... 36 5.3.1. Sonorants .................................................................................................................... 36 5.3.2. Obstruents .................................................................................................................. 41 5.4. Discussion of consonant inventories .................................................................................. 51 5.5. Stress and length ................................................................................................................ 58 5.5.1. Stress .......................................................................................................................... 58 5.5.2. Length in vowels ........................................................................................................ 59 5.5.3. Length in consonants ................................................................................................. 61 6. Discussion .................................................................................................................................. 63 7. Conclusion ................................................................................................................................. 70 References ..................................................................................................................................... 71 Appendix 1 – Sorting of wordlists ................................................................................................. 74 Appendix 2 – Cognates used in the analysis ................................................................................. 82 Appendix 3 – Correspondence sets ................................................................................................ 85 Appendix 4 – Protoforms ............................................................................................................... 96 1
  • 2. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. Abstract This paper investigates phonological contact effects through a case study of Finno-Romani. Existing research on the topic is presented, covering sociolinguistic, structural and typological studies. The development of Finno-Romani since its contact with Finnish is examined using the comparative method, and the changes are discussed in relation to the sound system of Finnish. I find that many changes are predictable from a combination of knowledge of the sociolinguistic situation and analysis of the sounds of the involved languages. In addition to this, I also find that certain changes in Finno-Romani cannot be explained solely with contact with Finnish, but that these changes have nevertheless happened because of it. I discuss two different views of phonological change and how they are applicable to the case of Finno-Romani, and find that although the two are not immediately reconcilable, they each contribute important information. I conclude that more studies with a structural focus are needed in order to better understand the mechanics of phonological change in contact situations. 1. Introduction In this thesis I systematically examine the process of phonetic and phonological change in a language in close contact with another language. The goal is to produce a thorough case-study of the phonetic and phonological aspects of language contact, and to propose future ways to study this phenomenon. The case-study in question is of Finno-Romani which has been in very close contact with Swedish and Finnish for a long time, and is reported to have changed significantly as a result of this contact, including its phonology (Granqvist 2002; Granqvist 1999a; Granqvist 1999b). Romani is well suited for this kind of study because its various branches have been in contact with many different languages. For this reason, there are rich grounds for future comparisons. Finno- Romani in particular is suited for this explorative case-study as its geographical placement means that it has a much more limited range of possible contact languages than for instance some of the Romani dialects of Central Europe. When reconstructing the development of Finno-Romani's sound system since its split from the other Romani dialects, I compare the changes in Finno-Romani with the phonological systems in its latest contact language, Finnish. This is in order to determine which of the possible contact-induced changes have happened and which have not, and what this might reveal about the mechanics of language change in contact situations. There are many different aspects of language contact and many different ways of studying 2
  • 3. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. them (sociolinguistic studies, studies of politically motivated language change, studies of mixed languages and studies of language attrition to name but a few), and all of these aspects are obviously interconnected and can each provide a deeper understanding of the others. In section 2 I provide a general overview of the suggested explanations of contact-induced language change that seem most relevant to the present study. However, as this subject field is so enormous, I focus in the main analysis almost exclusively on the structural aspects of language change and what these might reveal about the general workings of human languages. Languages are in close contact with other languages all over the world, and the situation in which they start to affect one another can reveal much about the nature of human language. From a structural point of view, any situation in which two language systems are in contact with each other can reveal something crucial about both the languages involved and the human language capacity. Or, as Matras (2009: 3) puts it, it can help us gain “an understanding of the inner functions and the inner structure of 'grammar' and the language faculty itself”. For this reason, as well as for many others, the field of contact linguistics is large and growing (Matras 2009: 3), but to the best of my knowledge no studies like the present one have been carried out: A study in which a careful mapping of the phonetic and phonological language history is compared to that of the given language's contact language(s). Two studies examine phonetic and phonological effects of language contact: Shaw & Balusu (2010) study changes in Japanese fricatives caused by contact with English, and Lleó, Cortès & Benet (2008) deal with changes in vowels in Catalan caused by contact with Spanish. Both do so by choosing a specific sound type and examining only that. In the case of Lleó, Cortès & Benet 2008, it is not even clear what the motivation for choosing those specific sound types is. Inspired by the point made by Thomason (2001: 93) that it is important to study the whole of a system when determining whether a change is contact-induced or not, I therefore examine the whole of the segmental inventories involved, as well as stress patterns and length in vowels and consonants. Ideally, the scope would be even broader and include not only phonotactics, but also morphological, syntactic and sociolinguistic factors as well, in addition to more precise phonetic measurements. This would provide a larger context for the sound changes in language contact studies. However, given the limited space available for this study, I have deemed it more useful to pick out a single, clearly demarcated subject of analysis instead of including bits and pieces from all domains. The term 'language contact' covers a broad spectrum of different kinds and levels of contact. My focus in this paper is on situations in which one language has caused considerable change in 3
  • 4. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. another but not (at least not yet) caused actual language death, and in which the changed language can still be recognised as the original language. That is, I will not focus on actual mixed languages or creoles, nor on language attrition or substratum effects, only on intensive contact in a situation of bilingualism. As already mentioned, my focus within the area of language contact is phonetics and phonology. There are several reasons for this. First and foremost, it is a field of study which – compared to contact morphology and syntax – has been neglected in recent years. Second, given that phonetics and subsequently phonology work within the boundaries of the human articulatory capacity, it offers a more finite set of variations than morphology and syntax, which are limited mainly by cognition. This limitation makes it easier to gain a good understanding of which contact effects are more likely to occur as many can be ruled out from the very beginning. Given the focus on the structural aspects and the declared goal of investigating the human language faculty through it, I will also touch (however briefly) upon the psycholinguistic aspects of bilingualism and multilingualism, as it is neither entirely possible nor indeed desirable to separate the synchronic aspects of language change from the diachronic aspects. In this I am inspired by Matras (2009: 2-6), whose declared goal it is to incorporate synchronic as well as diachronic studies into a coherent, functionalist model of communication which also takes into account “inner coherence of language systems” as well as models of bilingual language processing, typological findings and contact linguistics. While the construction of such a model is very much outside the scope of this study, I nevertheless present my analysis in a way that might later be useful in the discussion of such models. This is clearly an area of study which is nowhere near exhaustion and one which I hope to contribute to. The structure of this paper is as follows: In section 2 I give a more detailed presentation of the work already done in the field of contact linguistics in general and in phonetic and phonological aspects of language contact in particular. I then explain my choice of method and the data used in the analysis (section 3), and then move on to the analysis of my chosen case: Finno-Romani (section 4). This is followed by an actual analysis of the sound changes that have taken place in Finno-Romani since its arrival in Finland, and these are compared with Finnish throughout the analysis (section 5). I round off this presentation with a discussion of my findings and the theoretical consequences they have for the study of phonetics and phonology in contact situations (section 6) and I conclude the paper in section 7. 4
  • 5. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. 2. Background I now present the research area of contact-induced sound change. In order to make the presentation more structured, I split the topic into three different branches, following Matras 2007. Each of these branches will be presented and related to the subject of the case-study presented in section 5. In an overview of proposed explanations of structural borrowing, Matras (2007: 34) identifies three different directions that explanations of contact-induced change might take: (1) Explanations that are concerned with “the degree of borrowing as related to the intensity of exposure to the contact language” (from here called 'sociolinguistic explanations'), (2) Explanations that are concerned with “the outcome of language contact as a product of the structural similarities and differences (congruence) among the languages concerned” (called 'structural explanations'), and (3) Explanations that are concerned with borrowability as a product of inherent traits of a given category (called 'typological explanations'). As all three aspects (sociolinguistic, structural and typological) are obviously relevant to an explanation of a contact situation and its effects, I will not in the following limit myself to dealing with only one of them. In fact, they are not entirely separable anyway. The main analysis in this study, however, is designed to cast light on the process from the point of view of the second type of explanations – the structural. In order to give the reader a solid context for the structural analysis, I also give an overview of the first (sociolinguistic explanations) and third (typological explanations) types of explanations, and provide the most basic information needed to understand how Finno- Romani might be explained according to these. 2.1. Sociolinguistic explanations Without doubt the most influential work concerned with the degree of contact and its consequence for the degree of borrowing, is the monograph by Thomason & Kaufman (1988). As is very clearly stated, Thomason & Kaufman hold the view that, while purely linguistic considerations are relevant, “it is the sociolinguistic history of the speakers, and not the structure of their language, that is the primary determinant of the linguistic outcome of language contact” (Thomason & Kaufman 1988: 35). While it has been suggested that this is not quite right (Treffers-Daller 1999), the view has also been defended from many different sides (see for instance Beeching 1999; Poplack & Meechan 1999; Sebba 1999; Thomason 1999; Winford 1999). To my knowledge nobody disputes that the degree and type of contact is indeed very relevant to the linguistic outcome of the 5
  • 6. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. contact situation. In the frame-work of Thomason & Kaufman (1988: 50), the sociolinguistic circumstances of Finno-Romani can be characterised as being a case of language maintenance (as opposed to language shift) as Finno-Romani is not quite in the process of being abandoned by their speakers (see Vuorela & Borin 1998 for a more nuanced view on this though). In addition to this, it is a case with “intensive contact, including much bilingualism among borrowing-language speakers over a long period of time” (Thomason & Kaufman 1988: 50). Indeed, speakers of Finno-Romani are probably all dominant in Finnish. According to the model, this leads to a high degree of lexical borrowing, and moderate to heavy structural borrowing, especially at the phonological domain. This is to some extent true, although the lexical borrowing from Finnish into Finno-Romani is surprisingly limited (for more on this, see section 3.2). Finally, Finno-Romani can be argued to be under a fairly strong cultural pressure from the surrounding Finnish-speaking society (see however again Vuorela & Borin 1998 for a more nuanced presentation of the situation), which is predicted to lead to massive grammatical replacement. This is also true, depending on the exact definition of 'massive'. The model thus seems to predict the results of the contact quite well based on the sociolinguistic parameters of the situation. In a similar tradition, but within a more speaker and communication oriented model, Matras 2009 proposes a functionalist view of communication in contact situations which explains contact- induced language change with the bilingual speaker as the central actor. In this model, change in contact situations happens to balance two opposing tendencies: The tendency to remain loyal to a specific repertoire in specific contexts (that is – to keep the acquired languages strictly separated and to keep each language linked to its specific social contexts) and the tendency to exploit the full expressive potential of the whole of the available repertoire (that is – to mix the acquired languages in order to be able to choose the most fitting trait in the situation, no matter which language specific items belong to). The way to balance these things is to reduce the amount of choices the bilingual has to make between the two opposing tendencies. This balance can be achieved for the speaker by making the involved languages more similar to each other. If one language borrows a word from the other, it becomes unnecessary for the bilingual speaker to make a decision of whether or not to use the given word in a context that invites loyalty to the recipient language. At the same time, the full expressive potential is freely available. When it comes to phonology in a contact situation specifically, Matras (2009: 221) notes that certain aspects are notably different from other grammatical domains such as morphology and syntax. Firstly, sound material falls somewhere between matter (content) and pattern (structure; for 6
  • 7. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. a more thorough discussion of these terms and their roles in language change, see Sakel 2007). Sounds obviously have a clear and audible content (and therefore matter), but at the same time sounds on their own cannot carry actual meaning in the way that for instance morphemes can. Therefore, the borrowing of sound patterns is not quite comparable to the borrowing of grammatical structures. Secondly, physiology restrains phonetics and phonology, but not morphology and syntax. There are therefore limits to the speakers' choice of whether or not to adopt new sounds from a contact language, especially if the speaker is an adult. Matras (2009) claims that all levels of phonetics and phonology may be borrowed, from phonetic features to phonemes to more prosodic features. Based on various degrees of contact (and in this his model is similar to that of Thomason & Kaufman 1988), he then lists four different types of processes leading to contact-induced phonological change. These four types (called A, B, C and D) cover both changes in a speaker's native language and in the contact language as spoken by the bilingual non-native speaker. In type “A” changes happen to elements borrowed into a language and not to the language that borrowed them. This is the kind of change that Singh (1996) calls “adaptation” and in which the phonology of borrowed words are adapted to fit the phonology of the recipient language. In Matras' speaker-centered model, this reflects a situation in which the speakers wish to remain loyal to their native language (while obviously still exploiting the potential for expression, given the fact that the borrowing has happened at all). In the type called “B”, there is also borrowing of words from a contact language into the recipient language, but no phonological adaptation is taking place. Rather, the borrowed words are allowed to keep their original shape leading to an enrichment of the phonology of the recipient language. This is what Singh (1996) calls “incorporation”, and this requires a degree of competence in the bilingual speakers which is not necessary for adaptation. In Matras' model, this reflects a situation in which the speakers are less concerned with staying loyal to their native language, and they are more concerned with allowing borrowed words to keep their foreign traits. This is generally due to the source language having a higher status than the recipient language. Both type A and type B describe situations in which individual borrowed words are affected and not the whole phonological system. In the type called “C”, there is widespread bilingualism, and speakers are uncomfortable maintaining two separate phonological systems. The solution to this for the speakers is to speak the contact language with something close to, or significantly influenced by, the native language's phonology. This is frequently described as 'substrate effects' and in individuals the process is 7
  • 8. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. commonly known simply as 'accent': The speakers are competent in both languages but find it easier to use the same phonological system (or at least parts of it) in both. In Matras' model this helps speakers avoid the choice between which phonological system is most fitting in a given context. One might argue that if the language spoken has to be decided upon, then the choice of phonology is a simple consequence of this choice and should therefore not offer any dilemma. Matras does not specifically comment on this, but, as already mentioned, it is noted that there are physiological limitations as well as communicative ones when it comes to phonology. I find it likely that physiological ability is more relevant to the phonological aspect than the competing forces of loyalty and communicative efficiency. Type C is similar to type A in that there is a transfer of the native language's features onto the foreign material, and not the other way around. In the type called “D”, something similar to type C happens, but in this case the contact language is dominant enough (that is, it has high status and there are many and competent bilinguals), that it is the phonology of the native language ('native' in this case being a somewhat dubious label as it might not always be the speaker's first language) which is adapted to fit the phonology of the contact language. Type D is similar to type B in that native material adapts to fit the foreign sound patterns. Type D is similar to type C in that speakers in both of these types try to avoid the selection of sounds from two distinct inventory sets and therefore end up using just one set. There is no claim that the four types are meant as each describing only one specific and distinct type of language contact, as there is overlap between the types. Furthermore, various types can often be found in the same contact situation. The main claim is that speakers in contact situations constantly navigate the opposing pressures of loyalty to one given language and the practical urge not to have to choose between two separate sound systems. The result is that speakers find themselves somewhere on a continuum which goes from having two completely separate systems and having the systems overlap and essentially become one. The placement on this continuum correlates with the degree of bilingualism found in the community: The higher the level of bilingualism, the more likely a convergence of the involved systems is found. While it is outside of the scope of this study to discuss sociolinguistic factors thoroughly, I will in this section - drawing heavily on work done by Vuorela & Borin 1998 - briefly summarise what is known about the social and political conditions of Finno-Romani in present and past. This is intended to give the reader an idea about which sort of language contact is being dealt with here, and where contact effects might have come from. 8
  • 9. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. The Roma-speakers of Finland are descendants of a group of Roma who came to Sweden in the early 16th century, after having split from the other Roma groups (which are now spread across Europe). They first appeared in Finland (which was a part of Sweden until 1809) in the late 16th century (Vuorela & Borin 1998: 52). The Roma continued to come to Finland from Sweden until 1809, when Finland became Russian, and the Roma residing in Finland at that time are thought to be the ancestors of the modern day Finnish Roma (Vuorela & Borin 1998: 53). To this day, many Finnish Roma (an estimated third of the total 9000) live partly in Finland and partly in Sweden (Vuorela & Borin 1998: 55). During the centuries since the Roma's arrival in Finland, they have been singled out as a group with various negative traits by authorities (Vuorela & Borin 1998: 54). and for this reason Finno-Romani has always been a low-status language. All Finnish Roma also speak Finnish and for many of them it is their first and perhaps even only language. Nevertheless, Romani is kept alive as an in-group language which is learnt in late childhood as a part of the adoption of Roma cultural behaviour (Vuorela & Borin 1998: 60). In 1998 Finland signed the European Charter for Regional and Minority Languages and identified Romani as a non-territorial minority language of Finland (Ministry of Social Affairs and Health of Finland 2004: 22). This however does not mean that the Roma (and in extension their language) generally enjoy a high status. All speakers of Finno-Romani are bilingual, and as already mentioned, Romani is in most families learned in late childhood. Romani is therefore the second language of the younger generations with Finnish being the first. According to the models of both Thomason & Kaufman (1988) and Matras (2009), this puts the contact situation between Finnish and Romani in the groups that is expected to show the most contact effects. According to Matras' model, given this degree of contact, the two phonological systems are likely to have more or less converged into one. Whether or not this expectation holds true will be discussed further in section 6. 2.2. Structural explanations When it comes to the more structurally-oriented explanations of the processes of contact-induced phonological change, there is to my knowledge no coherent model of what might be the expected outcome when two languages are in close contact. However, there are several smaller contributions to the field, and I will present some of them here. Still within the framework of the model described above, Matras (2009) describes various types of phonological contact effects based primarily on the empirical work described in (Matras 9
  • 10. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. 2007). As these various effects do not seem to be grouped or in any hierarchy, they will simply be presented one by one. One process that is commonly found in situations with a great amount of contact, is that sounds in loanwords are reinterpreted as native sounds. Matras (2009: 226) notes the following of the process: “Speakers perceive similarities between a sound X in one language, and a sound Y in another. These similarities usually derive from shared features in the position of articulation of the two sounds, sometimes also in the mode of articulation. On this basis, one sound is allowed to represent the other.” In types A and C this process happens with the native language as the template and the borrowed words as something which is mapped onto it. This process is called 'approximation'. Another process is the loss of features, and this process leads to the omission of one or more features in a phoneme in a loanword. The omission is typically due to the feature being difficult for speakers of the recipient language to pronounce, and omissions are therefore unlikely to have counterparts in the recipient language. This might happen not only to single features but also to phonemes or specific phoneme combinations such as consonant clusters, which might be reduced in loanwords. In some cases these reductions of features or phonemes may result in a disruption of the general phonological system, as some distinctions might be lost. This might lead to words which were minimal pairs in the source language to merge in the recipient language. Related to feature loss is phoneme substitution. In this process, phonemes in loanwords are substituted by phonemes in the recipient language that share enough features with them, so that they seem somewhat similar. According to Matras (2009: 227), this is a very common process, as it is very likely that a given pair of languages share features of a given phoneme but a lot more unlikely that they share the exact same combination of features in the same phoneme. Syllable structure and stress systems too are susceptible to changes when loanwords are incorporated into a recipient language, although this does not always happen. Even when no actual changes in the phonological system happen, there might be a frequency effect when a certain phoneme with a somewhat peripheral function in the recipient language is given a more prominent role due to its occurrence in many loanwords. It might also happen that loanwords are allowed a life outside the general phonological rules of the recipient language. In contrast to this it might also happen that an extensive borrowing leads to a deeper change in the phonology than merely an enrichment of the inventory. If, for instance, a sufficiently large amount of loanwords are borrowed in which a given phoneme type is found in positions where it is not usually found in the recipient 10
  • 11. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. language, then the established system might adjust. In cases with even more effect on the recipient language's phonology (“convergence”, Matras 2009: 229), the similarities between the languages involved increase. This might lead to enrichment of one of the inventories, but also loss of phonemes in one of the languages. A very frequent stating point of convergence is the borrowing of an allophonic variation and/or the rules for its distribution. In direct relation to the notion that phonological convergence starts with the borrowing of allophonic variations or distributions rather than of new sounds, Shaw & Balusu (2010) present a study of changes found in Japanese due to contact with English. Their proposal is that new phonological distinctions that are found in loans, are parasitic on existing phonetic (but non- phonemic) differences found in the receiving language. The example studied by Shaw & Balusu (2010) is the phonological contrast between [ti] and [tʃi] which is found in Japanese in loanwords from English. Japanese does have the contrast [ti] – [tʃi], but only as a prosodically conditioned difference: [tʃi] appears in prosodically strong environment such as heads of feet and accented syllables, while [ti] appears elsewhere. The phonological contrast in the English loanwords thus has a ready phonetic continuum to map unto, and it is the suggestion of Shaw & Balusu (2010: 157) that this is a prerequisite for the preservation of the contrast in the loanwords. The suggestion is strengthened by the reported fact that most contrasts not found in the native phonology are simply neutralised in English loanwords in Japanese. Singh (1996) combines structural explanations with typological ones to explain why certain features will be more easily borrowed than others. Citing The Universal Syllabic Template (which states that the nucleus of a syllable is the peak of sonority and that going out from this peak (to both sides) the sonority will be steadily falling), he states that it is easy to borrow something that fits a general principle (such as the Universal Syllable Template) even if the recipient language originally has a gap in that place. As an example he mentions that English does not have the consonant combination [vl] in syllable onset even though this fits the Universal Syllable Template. However, the fact that it does fit the Universal Syllable Template, makes it easy for English speakers to pronounce borrowed words from Russian such as 'Vladimir', and this fact will make words with these phoneme sequences more likely to be borrowed unchanged into English than others. Lleó, Cortès & Benet (2008) set up two different sets of expectations to the outcome of a study of Catalan vowels during contact with Spanish. One of the sets of expectations is based on a general typological basis while the other is based on a comparison of Spanish and Catalan phonologies. While the results are not entirely clear-cut, the comparison of the specific phonologies 11
  • 12. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. is deemed by the authors to be the best predictor of the outcome. Catalan has a vowel system with four mid-vowels: /e/, /o/, /ɛ/ and /ɔ/ whereas Spanish only has /e/ and /o/. As the Spanish system is less complex, it is expected that Catalan in contact with Spanish will have a tendency to merge /e/ with /ɛ/ and /o/ with /ɔ/ (cf. Matras' description of phoneme loss in section 2.2). The authors wonder which of the two original mid-vowel sets ([e] – [o] and [ɛ] - [ɔ]) will be the realisation of this merged phone in Catalan. In order to investigate this, they make recordings of three generations of speakers of Catalan living in two different parts of Barcelona with different proportions of Catalan speakers. It turns out that in speakers who displays most contact effects, the [e] - [o] vowel set is the one primarily found in Catalan as the realisations of the merged phonemes. Lleó, Cortès & Benet (2008) do not really explain why they chose to look at vowels rather than something else, or how they started suspecting that there might be something to look for. I find that this lack of insight into the whole of the phonological systems makes the study a little hard to assess. As structural explanations of contact-induced change are the main focus of sections 3 - 6 and as they will be treated there, I will not elaborate on the phonological structure of Finno-Romani or Finnish here. 2.3. Typological explanations Explanations of phonological contact-induced changes based on typological factors are relatively rare, and the reason seems to be that there simply are not very many: “it does not seem possible at this stage to point to any position within the phonological system (e.g. certain articulatory modes or positions, marked features, etc.) as being particularly prone to contact-induced change. It seems that the details of phonological change are entirely a product of the relations among the two systems – or congruence – and any statistics of change are likely to simply reflect the mere likelihood of the two phoneme systems in contact to share certain phonemes, and to differ with respect of others.” (Matras 2007: 39). Matras (2009: 228) does add that new consonants are more likely to be adopted than new vowels, but this is attributed to the fact that consonant inventories are generally larger than vowel inventories and that there is therefore a higher potential among the consonants of not overlapping between the involved languages. This entails a higher pressure on the consonant systems to adjust than the vowel systems. Also, prosodic features seem to be more susceptible to borrowing than segmental phonology, which is attributed partly to the somewhat peripheral role prosody has to meaning and partly to “the proven neurophysical separation between prosody and 12
  • 13. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. other aspects of speech production” (Matras 2009: 233). Here it is quite relevant to mention that Matras (2009: 231) does not seem to count stress as a prosodic feature stating “prosody seems to be more prone to cross-linguistic replication in contact situations than segmental phonology, with stress figuring in-between the two”. It has been proposed that markedness is highly relevant for the likelihood a trait has of being transferred. Lleó, Cortès & Benet (2008) present a “Jakobsonian perspective” on markedness (for more on this, see Jakobson 1968) in which “marked entities presuppose unmarked ones in a typological, diachronic and acquisition sense” (Lleó, Cortès & Benet 2008: 188), meaning that for a language to have a marked trait, it must also have the unmarked corresponding version of it. It is never made entirely clear by Lleó, Cortès & Benet (2008) what exactly one should understand by 'corresponding'. In the case of phonology it seems clear enough that a voiced stop can be said to be more marked than an unvoiced one with the same place of articulation, but how far the notion of 'correspondence' can be taken is unclear. Is an unvoiced fricative in any way 'corresponding' to an unvoiced stop, for instance? Another factor which Lleó, Cortès & Benet (2008: 188) propose as relevant to borrowability is complexity. 'Complexity' is meant as a psycholinguistic notion and is specifically linked to allophony. The more allophones a phoneme has, the more complex it will be, as it will cost the learner more work: There is more decoding to be done, and also more rules to learn in order to produce the right phone in the right environment. Although it is never overtly stated, the assumption seems to be that a complex trait might be harder to borrow than a simple one. This assumes that phonemes are borrowed as a complete packet with all the allophones of the source language. Finally, Lleó, Cortès & Benet (2008: 188) also suggest that traits which are frequent and dominating in the source language are more easily transferred to another language. As Lleó, Cortès & Benet (2008: 188) point out, this should not be confused with frequency in all the languages of the world as that would be a matter of markedness rather than frequency. This suggestion is not really followed up in the discussion of the results, but the overall conclusion is that a comparison of the involved phonological systems (as described in section 2.2) is a better predictor of the outcome of language contact than any general, inherent traits of the constituents of the phonologies. 2.4. Summary In conclusion, sociolinguistic explanations of situations of language contact can successfully predict the degree of contact effects found in the affected language(s). However, if one wants to know more 13
  • 14. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. about the contact effects than merely their degree, other explanations and models are necessary. While typological explanations have limited success in predicting the outcome of language contact, more structural approaches do seem capable of predicting not only the degree but also more specifically the type of contact effects. Still, there are limited resources on this in the existing literature. Shaw & Balusu (2010) and Lleó, Cortès & Benet (2008) each cover only small corners of the phonological systems of Japanese and Catalan respectively, and while Matras (2009) does list many possible effects, he does not really comment on how these might be interrelated in one phonological system. The overall focus missing thus seems to be a cohesive investigation of whole, specific cases. In order to at least begin to fill this gap, I now turn to an analysis of Finno-Romani and the phonetic and phonological changes which have happened to it since it became a separate dialect branch. Finno-Romani is very useful for investigations of this sort. Firstly, various branches of Romani have been in contact with so many different languages that there are very rich possibilities for further comparisons. These comparisons might be studies of how similar languages behave in contact with different languages – something that might give a greater understanding of what language is and needs in order to be a language. This makes Romani in general uniquely suited to this sort of study. Secondly, Finno-Romani is a collection of very similar but not quite identical dialect variations. It is found in a relatively isolated part of Europe as Finland is geographically at the fringe of the European landmass and also to a great extent surrounded by water, and addition to this, speakers of Finno-Romani are rarely in contact with speakers of other Romani varieties. These circumstances makes it, in a way, simply easier to work with than many other Romani dialects spoken in for example Central Europe because there are limited sources of contact. I therefore deem Finno-Romani a good starting point, and in the following I outline my approach to its investigation. 3. Method In this section I describe the data I use and the data-base they are collected from (section 3.1) and outline the procedure of the analysis to be described in sections 4 and 5 which is the basis of the results summarised and discussed in 6 (3.2). This section also includes considerations about possible but rejected procedures and the evolution of the remaining one, as it has changed a great deal since the beginning of this project, and with good reason. I briefly describe the reasons for 14
  • 15. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. these changes and why I find the present approach the most reasonable. 3.1. Data The data used in this thesis is all collected from the Romani Morpho-Syntax database (RMS- database, http://romani.humanities.manchester.ac.uk/rms/), (see Matras, White & Elšik 2009 for a more detailed description). The RMS-database is a publicly accessible online collection of sound files and transcriptions of speakers of various Romani-dialects reading aloud wordlists, translating sentences into Romani, and engaging in narration. In addition to being simply a collection, the database also makes various modes of cross-dialectic search available and is thus a very useful tool in many different kinds of studies involving Romani (see for instance Elšik & Matras 2006) who used it to compare morphological and syntactic changes found in the various dialects with the languages they had each been in contact with). In most cases, the RMS-database does not give information about their transcribers, but they are described as being mainly “graduate students specialising in Romani linguistics” (Matras, White & Elšik 2009: 11). It is further noted that a great amount of work went into assuring that the transcribers followed the same standards and conventions for the transcription. It is also mentioned that the collection in the RMS-database was gathered using an international network. I take this to mean that the transcribers working on Finno- Romani were probably living in Finland as well and presumably native speakers of Finnish (which they could be assumed to be even if they happened to be Finnish Roma, as mentioned in section 2.1). The transcription conventions used in the RMS-database are based on a standard transcription consensus of Romani, which I have “translated” into broad IPA using both the sound files attached to each word in the database and the existing transcriptions in the database. I found this the best approach as I am well aware that there are important differences between various phonemes which I being a Dane - albeit a linguistically trained one – would be hard pressed to notice, but which are easily detected by native speakers of Finnish, especially in the realm of allophones. As several sources note that Finno-Romani sounds a lot like Finnish (Granqvist 2002; Vuorela & Borin 1998), it seemed the most prudent not to dismiss distinctions made by speakers of Finnish even in cases when I myself could not hear it. In a very few cases I have deviated from the existing transcriptions, as differences in the existing transcription and my own interpretation of the pronunciation seemed large enough to justify it. These cases are discussed in the analysis. A last note on the transcription is that as stress almost always falls on the first syllable in Finno-Romani, stress is marked only 15
  • 16. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. when this is not the case. When it comes to Finno-Romani, the database provides word lists from six different speakers. Two of these are noted as speaking West-Finnish Romani (These are referred to as FIN- 005 and FIN-011) and the remaining four East-Finnish (Referred to as FIN-002, FIN-006, FIN-008 and FIN-012). One of the West-Finnish lists (FIN-011) is very incomplete and has been excluded from the analysis for this reason. One of the East-Finnish lists (FIN-002) does not have corresponding sound files which makes it impossible to verify the transcription given, but as the transcriptions of the other dialects are generally good, I have chosen to include this sample. The sound files of FIN-008 have a lot of background noise which makes it virtually impossible to judge the transcription of especially fricatives and affricates, but this sample has been included in the analysis too for the same reason that FIN-002 has. In the following, the different samples will be treated as different dialects although there is only one speaker for each sample. It is quite possible that others factors such as speaker age and the register used have a greater influence on the data than the geographical placement and dialect variety does. However, as I am investigating different outcomes of Finno-Romani in contact with Finnish, it is not so very important which of the factors might be most influential. In addition to the samples of Finnish Romani, I also use the two Sinti dialects available in the database: One currently spoken in Romania (RO-022), which has until very recently been spoken in Germany, and another dialect currently spoken in Austria (AT-001x). These are assumed to belong to the group called German Sinti by Matras (2002: 9). I include these samples as Sinti is closely related to Finno-Romani. Together they form the subgroup Northwestern Romani and they have many lexical parallels (Boretzky 2012), which makes them suitable for an analysis and reconstruction like the one described in section 3.2. Using a resource like the RMS-database has a number of advantages. Most notably, it saves the researcher doing an analysis very much time, energy and money not having to collect their own data. In a collection like this, much work has also been done so thoroughly and with so competent people involved that the quality of the data is probably better than what one person could collect within a reasonable time-span. That said, there are of course also disadvantages: Mainly, that one does not get to choose which information is relevant. I would have preferred to know if all the samples were indeed transcribed by native speakers of Finnish as I have assumed, and also what age the informants had and what their background was. I find that the advantages are greater than the disadvantages, and so I judge it reasonable to use the RMS-database for this investigation. 16
  • 17. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. 3.2. Procedure I now describe the procedure I use to get the results presented in this project. Briefly stated, the goal is to determine which sound changes have happened in Finno-Romani since its split from Sinti and other related varieties, and consequent close contact with (Swedish and) Finnish. These changes are then compared to the Finnish sounds and sound structures in an attempt to determine whether or not and to what extent the changes are due to contact with Finnish. The procedure has changed somewhat in the course of the project, and as these changes are very much a part of the process of working with the data, and as they are important to know in order to understand the reasoning behind my method, I will describe those as well. Using the data from the RMS-database, the goal is to determine the course of phonetic and phonological changes in Finno-Romani and whether or not these can be ascribed to contact effects. The first part of the procedure is to determine which words in the word-lists are inherited and which ones are borrowed. The inherited words are needed for input to the comparative method (described below) and this is made difficult if recently borrowed material is used, so this distinction is quite relevant. Originally, the intention was to also use the borrowed words because I assumed that most of them would be from Finnish. The idea was to examine these loanwords in order to make a description of the phonetic and phonological systems they represented when they were borrowed into Romani. These systems were then expected to be the systems of Finnish at the time of the adoption. While the form that was borrowed into Romani might very well have been adapted already at the point of adoption, this was not considered to be a problem as I was happy to investigate the part of the process taking place after the adoption. The process of adopting loanwords is a highly interesting topic, but it is more easily examined in studies with a more synchronic approach such as Lleó, Cortès & Benet (2008) and Shaw & Balusu (2010). I considered the diachronic perspective of this study to be much more appropriate to investigate the process happening after the adaptation (although this approach did assume that the adaptation of the words to the Romani phonology was not complete, and that there would be structural differences between the loanwords and the inherited words). My intention was then to compare the description of the phonetic and phonological properties of the loanwords to the phonetic and phonological properties of Romani and also to the changes which have happened in Finno-Romani since then. It became apparent that this method would not yield any results when the data sets made it clear that there are in fact very few loanwords from Finnish in Finno-Romani. The reason seems to be that Romani has 17
  • 18. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. been used as a secret language by the Finnish Roma (Vuorela & Borin 1998: 73), and this would obviously be less effective if too many of the words were recognisable to Finnish speakers. While this plan was then abandoned, I still needed to know what Romani had looked like immediately before contact with Finnish (or at least as close to 'immediately before' as it was possible to get) and which sound changes had happened since then. Inherited words have been separated from loanwords by a combination of comparing them with the other Romani dialects in the RMS-database and finding etymological information in the relevant literature (mainly Boretzky 2012; Valtonen 1972 and Matras 2010). If a word appears to have clear cognates in other, more remotely related dialects attested in the RMS-database, it is assumed to be an inherited word. The term “inherited” is somewhat relative in the case of Romani as many of the words assumed to be inherited with the described procedure will in fact be loanwords borrowed at an earlier stage in the history of the language, and potentially present in (virtually) all dialects or branches. Romani has borrowed extensively from other languages since its speakers' exodus from India (see for instance Matras 2002), and even core parts of the lexicon are therefore borrowed. However, in cases where a word is found to have cognates in different dialect branches, it is assumed to have been borrowed into Romani such a long time ago that is has been fully adopted to Romani phonology, and is therefore counted as an inherited word in this study. All of these observations are checked against the etymological information in the mentioned literature. Some of the more difficult cases of this procedure are described in section 4.1. The second part of the process is then to analyse the inherited words in order to establish the phonetics and phonology of Finno-Romani at the stage before it started adopting loanwords. “Loanword” in this context refers only to words which are not found to have any plausible cognates outside of Finno-Romani and which are not traceable as older loans. I do this by comparing Finno- Romani with the two Sinti dialects found in the database: RO-022 and AT-001x. Using the comparative method I determine the phonetic and phonological properties of the language stage immediately before Sinti and Finno-Romani split into separate dialects (this stage will from now on be referred to as 'Proto-Northwestern'). Proto-Northwestern is the closest it is possible to get to the phonetics and phonology of Finno-Romani immediately before adoption of loanwords and extensive contact with Swedish and Finnish. An easier solution would have been to simply use Early Romani, which is the (already) reconstructed language stage of Romani after extensive borrowing from Greek in somewhere around the tenth century (Elšik & Matras 2006: 68), but as I wanted to get as close to the stage of Proto-Northwestern as possible, I deemed this too imprecise 18
  • 19. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. and preferred to do my own reconstruction (I am also not aware of any actual word list with reconstructed forms). This procedure is further described in section 5. After having reconstructed the proto-stage and the sound changes which have happened since then, I compare the changes revealed by the analysis with the sounds and sound structure of Standard Spoken Finnish. Although the ideal subject of comparison would have been the Finnish dialects spoken in the places where the various Finno-Romani dialects in the analysis are found, I have chosen not to do so for several reasons. Firstly, very little material about the sounds and sound structure is available on these Finnish dialects (especially in languages other than Finnish), which would make for at comparison with great holes at best. Secondly, Standard Spoken Finnish is based on all the Finnish dialects and is not - like many other standard varieties one particular dialect which due to status has become the official language (Suomi, Toivanen & Ylitalo 2008: 7). Thirdly, a very thorough description of the Finnish sound system is available (Suomi, Toivanen & Ylitalo 2008), and it seems preferable to have as basis a seemingly complete and thorough view of the matter to a collection of partial sources which might not even agree. Fourthly, Suomi, Toivanen & Ylitalo (2008) actually do comment on dialectic variations, and these comments will of course be included whenever they occur and are deemed relevant. For these reasons I have decided that taking Standard Spoken Finnish as basis for comparison is the most reasonable approach to the matter. In many cases I use the forms of European Romani as listed by Matras (2010) for reference in the analysis. I have “translated” these forms into broad IPA in line with the other transcriptions in the paper, and they are listed in the final column in Appendix 2 for reference. Matras defines “European Romani” as “the reconstructed entity referred to as 'Early Romani' (Elšik & Matras 2006; Matras 2002), along with any modifications that can be assumed to have emerged in the particular forerunner dialect of British Romani. […] European loanwords that entered the language prior to the isolation of British Romani and which are widely attested in other dialects of the language are considered part of the European Romani legacy.” Although British Romani is an independent dialect branch (Matras 2002: 10), its forerunner – European Romani – is considered closely enough related to Finno-Romani to serve the present purpose. European Romani is a great help in reconstructing Proto-Northwestern, as it provides information about a language stage which comes before Proto-Northwestern and therefore makes it a lot easier to determine the direction of sound changes. While it is possible to make reconstructions without knowledge of such a language stage, it is more difficult and the results less certain. Given that the information is available, I believe it would have been waste of work and knowledge not to use it and to build upon it. The 19
  • 20. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. same goes for other published sources I use in the analysis: I see no reason not to make full use of them, as long as this is done with a critical eye. To my knowledge, no one has used exactly this procedure for a similar study. The comparative method is obviously well-used, but as far as I can tell not with this goal. Though there are aspects of this procedure which are perhaps not ideal, it is my claim that this is the best possible approach to this topic. 4. Preliminary data analysis In this section I give a detailed account of the first part of the analysis of Finno-Romani as described in section 3.2. This account shows the process of determining which words in the RMS- database' vocabulary lists of Finno-Romani are inherited and which are loanwords, and whatever else is needed before starting the main analysis. Appendix 1 shows the word lists for the Sinti dialects and Finno-Romani. Using existing literature (mainly Boretzky 2012 and Valtonen 1972) and comparisons with the other dialects in the RMS-database, all words in these lists have been sorted into three groups: words in multi-word phrases (written in grey in Appendix 1 and not included in further analysis), non-cognates (written in white on black and used sporadically for certain parts of the analysis and cognates (written in black on white and used for the central part of the analysis (section 5)). Several cases from the RMS-lists are not included in the table in Appendix 1. These are cases where no word form from any dialect was useful: Cases with autonyms and cases with only whole phrases instead of simple words. Compounds have also been excluded as these complicated the initial analysis. Many of the cases listed in Appendix 1 are further discussed in the following four sections: Cases whose status as loanwords or inherited words is doubtful (section 4.1), cases where it seems that there has been semantic shift (section 4.2), inflected words (section 4.4) and cases with more than one cognate (section 4.3). These cases are all marked with a letter in the column labelled 'comments' in Appendix 1. Cases where fewer than three dialects have inherited words (of the same cognacy) will not be used in the analysis in section 5 but are nevertheless shown in Appendix 1 with the relevant word forms noted as cognates. The outcome of this preliminary data analysis – a list with cognate sets ready for analysis – 20
  • 21. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. can be found in Appendix 2 and will work as the basis for the reconstruction described in section 5. 4.1. Cases which might be both loanwords and inherited words I now briefly discuss the cases in which there is doubt about whether a word form is a loanword or an inherited word. These cases are marked with the letter 'D' in the column 'Comments' in Appendix 1 and are discussed in numerical order here. 28 “cow” is a difficult case as the Finno-Romani forms [kurun], [kurni], [kurja], [gurunis] look like both the inherited 'gurumni' but also German 'Kuh' and Swedish 'ko' which both mean 'Cow'. Boretzky (2012: 23) lists the form 'gurni' (which but for the voicing of the initial plosive is identical with the forms of FIN-005 and FIN-008) as being a regular variation of the word, but only in South-Balkan dialects. The suffixes -in and -ja are both feminine (see also section 4.4) and go back to Sanskrit (-in, Valtonen 1972: 16). -ja is less certain, but probably a plural marker. It is however clear that at least -ja has been productive even after contact with Germanic languages, so the presence of old inflections does not rule out that these words might be loanwords. There is nothing to explain how the 'r' present in all dialects would have been added to 'Kuh' or 'ko' though, and for this reason, all the forms found in 28 “cow” have been counted as inherited words. Case number 70 “belly” is quite clearly a Germanic loanword (could be from German 'Magen' or Swedish 'mage' for instance), but as the word is found in both Sinti and Finno-Romani, it seems to have been borrowed before the two branches split, and it has therefore been included. the same goes for 81 “sky”. In 164 “lift” the forms [khantel] and [xandela] are found in FIN-002 and FIN-012 respectively. There is nothing to indicate that these forms are loanwords, and there is nothing in their phonological forms that does not fit with Finno-Romani, but on the other hand none of the other Romani dialects in the database have similar forms, and I have also not been able to find the form described anywhere. For this reason they are marked as loanwords, but as there are only two forms which might be useful, the case would not have been used in the analysis in any case. In 185 “meet”, the forms [tikkel] and [tikka:] do not seem to be the standard Romani form, but it turns out to be the word for “see”, which is case number 158. As the two are identical and 158 “see” is more complete, 185 “meet” is left out of the analysis. In 186 “come”, RO-022 has the form [va] which at first glance does not look like it is 21
  • 22. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. cognate with [a:vel] which is found in Finno-Romani. However, [avava] is very frequently found in the other Romani dialects in the RMS-database, and this form can very easily have both [a:vel] and [va] as its reflexes. I therefore keep both forms, but as Sinti is the dialects furthest from the most commonly found form, its differences will not be discussed further. Case number 197 “pay” is quite possibly not useful as a cognate. While there is a European Romani form ples-, there is even at this point doubt whether or not this is influenced by the German word for price, Preis (Matras 2010: 202). As this could have happened early enough to influence European Romani but also easily at the stage of Proto-Northwestern or even later, it is impossible to tell if the forms found in data reflect ples or preis and this case has therefore been excluded from the analysis. 4.2. Cases with semantic shift While it is outside the scope of this study to investigate semantic shifts in Romani, I will here briefly mention the cases where a shift seems the most likely explanation and describe how each case is handled in the analysis. Cases mentioned in this section are marked with the letter 'S' in the column 'Comments' in Appendix 1. In case number 46 “lip” FIN-002 has the form [mui] which is clearly an inherited word but which did not originally mean “lip” but rather “mouth”. It still means “mouth” in case number 47 in almost all the dialects, including FIN-002, which suggests that [mui] has undergone generalisation in this dialect. Another possible explanation is that something in the field situation (such as the researcher pointing) is responsible for this shift. As FIN-002 is the only dialect that uses [mui] to mean “lip”, and as there are thus no other reflexes of the cognate that mean “lip”, the [mui] in 46 will simply not be included in the segmental analysis. European Romani had two different words for “foot” (piro) and “leg” (heroj). Finno-Romani uses forms of piro for both meanings (see cases 61 and 65). In 61 “leg” RO-022 also has a form of piro, and the phrase RO-022 has in 65 “foot” also includes the word. This indicates that the same generalisation has partially taken place in Sinti (AT-001x has the form [heri] in 65, so the generalisation is clearly not complete in all the dialects), and it thus seems to be a generalisation which has happened before or at the stage of Proto-Northwestern and therefore outside the scope of this study. The two cases will be treated as two different cases as the different inflections of these cases in FIN-006 suggests that it might not be two completely identical cases. 22
  • 23. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. Matras (2010) lists no European Romani forms for “month” (84), but it is quite clear that all the dialects in this study have generalised the word for “moon” (83) to cover the meaning of “month” as well. As this does not seem to have happened after Proto-Northwestern, I will discuss it no further, but the two cases have been kept separate as they show slight differences in form. The word phuv meant “ground” in European Romani and not “earth” (which was the word chik) although it is used in this meaning in all the dialects in this study (see 87). This is not a new development and will not be treated further. In 91 “rain” FIN-002 uses the form [pa:ni] which is also found in 90 “water in both FIN-002 and the other Romani dialects. Given that [pa:ni] has a much wider use as the form for “water” than for “rain” in all the Romani dialects in the RMS-database, I assume that “water” is the original meaning. I simply leave FIN-002 out of the analysis of 90 “water”. The European Romani word mas meant “meat” while balavas meant “bacon”. While there is no word for “meat” in the wordlists in the RMS-database, all the Finno-Romani dialects use [mas] to mean “bacon” (case number 106). RO-022 has a completely different form in this case ([ʃuŋka]) which is from German Schinken, and it is possible that the change in meaning has happened only in Finno-Romani. As the only potential conflict is in the meaning though, and as bacon is meat, 106 “bacon” is treated as any other case. European Romani skamin meaning “chair” seems to be used in a generalised fashion in both Finno-Romani and AT-001x to also cover the meaning “bench” (case number 126) in addition to that of “chair” (case number 127). In both 126 and 127 RO-022 uses a different non-cognate which is inflected differently in the two cases and for this reason I am not comfortable lumping the two cases together even though the forms of skamin are identical. Sinti uses the word [la:p] for both “word” and “name”. It is originally the word for “word”, but its broader use is quite common among the Romani dialects (Boretzky 2012: 27). Sinti [la:p] is not included in the analysis of 135 “name”. In case 147 “light” Finno-Romani uses the form of the word which meant “fire” (see also case 95) in European Romani. A generalisation of this word has thus taken place, but apparently only in Finno-Romani. As 95 and 147 are not identical, they are treated separately. Cases 180 “love” and 181 “want” both go back to the European Romani word kam- which covered both meanings. It appears that the meaning of 182 “beg” has since been included as well, at 23
  • 24. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. least in Finno-Romani. As none of the three cases are identical inflection-wise, they have all been included in the analysis. 4.3. Cases with different cognates In this section I discuss some of the cases whose word forms reflect more than one cognate. These are all marked with the letter 'C' in Appendix 1. In cases where there are clearly two different cognates and at least one of them have fewer than three reflexes across the dialects, this cognate will not be used in the following analysis and will not figure in Appendix 2. For cases where this is the only comment, there will be no further mention. Case number 16 “horse” seems to have two quite different forms. Not only in Sinti and Finno-Romani, but also across the other Romani dialects (approximations to either 'gras' or 'graj'). It is of quite possible that the gras – graj difference has developed in Finno-Romani from a single ancestor, but the presence of both forms in many other dialects makes it more likely that there are two different cognates in game in Finno-Romani. As only FIN-008 has the form ending with a fricative, this occurrence will simply be taken out of the analysis and the other form will be analysed in the regular way. Already at the stage of European Romani the word “arm” had two different forms: musi and vast (Matras 2010: 189) with vast being the most general term meaning both “hand” and “arm” and musi being more specialised and meaning only “arm”. Various forms of vast still mean “hand” in all the dialects of this study, but FIN-005, FIN-008 and AT-001x use forms of musi for “arm” instead. Given that this division is found already in European Romani, it seems clear that this specialisation happened before the Proto-Northwestern stage and is therefore not the concern of this study. As there are three instances of both [vast] and [mus-] in 55, they are simply treated separately so that the forms of [vast] are called 55a and the forms of [mus-] 55b in Appendix 2. European Romani had a word for “tree” (rukh) and one for “wood” (kašt), but while all the dialects in the study still use forms of kašt to mean “wood”, FIN-005, FIN-006 and FIN-012 also use this to mean “tree” while the Sinti dialects and FIN-008 have kept forms of rukh. As there are three of each form in 96, this case has simply been split up into two separate cases: 96a and 96b. There are certain similarities in 146 “gold” between Sinti [symnokaj] and Finno-Romani [sonatiko] (and its variations), but they also look similar enough to possibly be cognates. European Romani has the form [sonakaj] which appears to be a mixture of the two forms. Both forms, and 24
  • 25. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. also forms that look like mixtures of the two, are found in the other Romani dialects in the RMS- database. It seems then that the two forms are cognates, but that the differences between the two go back much further than Proto-Northwestern, and I therefore treat them as different cognates. The same is true for 190 “leave sth”, in which [mukk-] and [mekk-] appear to reflect an older split (Boretzky 2012), for [arre] and [an] in 216 “in”, and for the Sinti [dis] in 77 “day”. In case number 199 “bad (nasul)” RO-022 [dʒungalo] appears similar enough to the Finno- Romani dialects ([pengalo]) that it seems fair to suggest similarity of form. However, as it turns out, beng is the word for “devil” in European Romani while džungalo originally meant “ugly”. The only similar thing is -alo, which is an adjectivising suffix. With those two cognates thus separated, there are not enough occurrences of the form to be included in the analysis. 4.4. Morphology In this section I make a few observations of inflectional and derivational morphology found in the inherited words in the data set. The aim is not to analyse the history of morphology in Finno- Romani in itself, and it is outside the scope of this study to attempt an explanation as to why which suffixes go on which words in which dialects. Instead, the goal is to use whatever morphological clues are available to further the phonological analysis, and most of all to avoid drawing conclusions about sound changes which are really a matter of different suffixes. All suffixes have therefore been removed both in cases where they were identical and in cases where they were different across the dialects. The morphemes themselves are not discussed here, but such discussion can be found by for instance Matras (2002). 4.5. Other considerations In a few cases there are circumstances to note which are not already presented. In 59 “finger” there are very clearly reflexes of the same cognate in all the dialects. However, these forms are unusually different compared to the rest of the cognates, which makes it almost impossible to know which segments should be matched up across the dialects. In order to keep the analysis as straight-forward as possible, I have only included the segments where there was no doubt in Appendix 3. In 92 “snow” the original RO-022 [ijp] from the RMS-database look like a transcription mistake. Judging by the sound file, there is nothing wrong with the transcription, but I find that [i:p] 25
  • 26. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. is a more reasonable interpretation than the original [ijp]. This makes the case much more straightforward to handle in the analysis. 5. Data analysis In this section I give a detailed account of the second part of the analysis of Finno-Romani as described in section 3.2, namely the analysis of the inherited words in Finno-Romani. The goal is to describe the language stage before Sinti and Finno-Romani split into different dialects (Proto- Northwestern), and the sound changes which have happened since then. I present the segmental inventory of Proto-Northwestern and the sound changes that have occurred since that stage (sections 5.1 and 5.3. In order to provide an idea of the context these segments are found in, I also briefly discuss stress and length (section 5.5). I first reconstruct and discuss the segmental inventory of Proto-Northwestern, which can be found in sections 5.1 and 5.3. To do so, I follow the procedure for using the comparative method as described by Crowley & Bowern (2010), starting with the setting up of correspondence sets for each cognate. An overview of all the correspondences can be found in Appendix 3. This is complicated by the many cases where one or more dialects for various reasons have gaps and therefore no correspondence. This sometimes makes it impossible to determine which type of correspondence a cognate set shows. Appendix 3 therefore simply shows all the correspondence sets with gaps, but the sets are grouped together according to the groups they most likely belong too (For instance, the set a;a;a;-;a;a;a most likely belongs to the set a;a;a;a;a;a;a as there are no obvious alternatives). In cases where there is more than one possibility, a set is grouped with other sets according to what seems most plausible, and this will be discussed in the following. Each group of correspondence sets in Appendix 3 is named for the protophoneme which I judge it to be reflexes of (left column). In cases where a protophoneme has sets of reflexes which are clearly different, these are grouped separately. All groups are also given a number in addition to the protophoneme for reference. One group of correspondences is marked with “N” as the relevant segments are different kinds of nasals – the N symbolizes an unspecified nasal consonant. Length is marked as a property of the vowels (using the standard IPA sign ” : ”) and long vowels are sorted together in the correspondence sets. This is done because the correspondence sets will serve as basis for the analysis and reconstruction of both the segmental inventory as well as stress and length. Terminology relating to language change are used as per Crowley & Bowern 2010. 26
  • 27. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. For each part of the analysis (vowel inventories, consonant inventories, length and stress), I compare the changes found with the sound structure of Standard Spoken Finnish (see 3) in an attempt to determine if or to which degree the change is motivated by contact with Finnish and also to discover which possible contact-induced changes have not taken place. It is outside the scope of this study to investigate phonotactics and syllable structure, but I use the following guideline in cases where it is useful to know the position of syllable boundaries: “every CV sequence is preceded by a syllable boundary” (Granqvist 2002: 74). 5.1. Analysis of correspondences with vowels In this section I present my reconstruction of the vowels of Proto-Northwestern and the changes which they have undergone since then. Firstly (this section), all the correspondences are presented briefly along with my analysis of them, and secondly (section 5.2), the results of the analysis are discussed and compared with Standard Spoken Finnish. Matters connected to length and stress as well as correspondences marked 'Diphthongs' are discussed separately in section 5.5. [a1-2+5-7] First of all, many words contain the sound [a] across all dialects. These appear to be unchanged since at least Proto-Northwestern and in many cases also since European Romani as they are also found there. This correspondence can be assumed to reflect *a. [a3+8] In three cases (186 “come”, 215 “outside” and 220 “we”), RO-022 has nothing, i.e zero (for instance in 215 [vri]) where the Finno-Romani dialects have [a] (for instance FIN-012 [auri]). This occurs in word-initial position in both cases. In 220 “we”, FIN-012 too does not have an [a] ([me]) while FIN-002, FIN-005 and FIN-008 do ([ame]). While RO-022 only rarely has [a] in word-initial position, it does occur (even among inherited words) and there is no clear system as to when an [a] disappears word-initially and when it does not. It is clear, however, from comparison with the other Romani dialects and European Romani that it is a case of apheresis (loss of an initial segment) in RO-022 and not of prothesis (insertion of an initial segment) in Finno-Romani. FIN-012 has word-initial [a]'s on several occasions, and there is no clear reason why it has been lost in 220 “we”. The correspondences of this group then also reflect *a. Although there are no clear conditions for the rule(s) that account for the change in RO-022 and FIN-012, the following rule does seem to have taken some effect in both 27
  • 28. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. dialects: *a → Ø / #_ (FIN-012 (Partial: 220) and RO-022 (Partial: 186, 215, 220)) [a4] In two cases (11 “girl” and 16 “horse”), the sound [æ] appears in some of the Finnish Romani- dialects where the rest of the dialects have [a] (for example FIN-005 [ʦæi] versus FIN-002 [ʦai] in 11). This looks like an example of assimilation (that one vowel has adopted one or more features – in this case it has become higher - from a neighbouring vowel) as [a] in these cases are next to [i], but as there are only two cases which are not even completely alike, it is not possible to set up a general rule for vowel assimilation in Finno-Romani. It does however seem clear that it can be found in FIN-005 and FIN-012. Although 82 (“star”) appears to be an exception as FIN-005 has the form [ʦerxai] and not [ʦerxæi], this can be explained by the fact that the diphthong [ai] in 82 (“star”) appears in an unstressed syllable. If the assimilation appears only in stressed syllables, there are no counter examples. Although FIN-008 only has a single case (11 “girl”) of vowel harmony, it too seems to have undergone a similar sound change, although it is found next to [j] rather than [i] in this case. Given how closely [i] and [j] resemble each other articulation-wise, it seems reasonable to combine them into the same rule. FIN-006 too seems to have undergone this change judging by 11(“girl”), but 16 (“horse”) is a clear counter example ([trai]). As there is no logical reason for 16 (“horse”) to have [a] and not [æ] (or indeed for 11 (“girl”) to have [æ] rather than [a] given that there are only the two cases where the rule might potentially apply), it is not possible to determine the exact course of change in FIN-006. It does however seem clear that these correspondences reflect *a as well. The final rule then looks like the following: *a [+stress] → æ / _ i,j (FIN-005 (Complete), FIN-006 (Partial: 11) FIN-008 (Partial: 11), FIN-012 (Partial: 11, 16)) [u1+3] When it comes to correspondences with [u], there are a few cases with plain [u]'s across all the dialects. There are not nearly as many of these as there are cases with plain [a] across all the dialects, but it still seems reasonable to reconstruct a *u for Proto-Northwestern for these cases and not note any changes. 28
  • 29. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. [v1] Another group consists of cases which have [u] in the Finno-Romani dialects and [p] in RO- 022 - for instance in case number 132 (“word”) which is [lau] in FIN-012 and [la:p] in RO-022. A comparison with all the other dialects in the database shows that the most widespread form of the [u]/[p] segment is [v] (which is also found in European Romani), and it seems plausible that this was also the case in Proto-Northwestern. Since then *v must then have undergone rounding and become syllabic in Finno-Romani ([u]) and have hardened to a plosive in RO-022 ([p]). This only happens when the original *v occurs in syllable-final position as shown by 186 “come” where FIN- 006 has [aulo] while FIN-005 has [a:vel]. The syllable boundary is assumed to be between [u] and [l] in [au.lo] and between [a:] and [v] in [a:.vel]. While this mostly happens after [a], 92 (“snow”) shows that it can also happen after [i], and 87 “earth” shows that it can also happen after [u] in which case [u] simply becomes long in Finno-Romani. This then seems to be a general rule that states that any *v in syllable-coda becomes [u] in Finno-Romani. The result is two new diphthongs, [au] and [iu], and lengthening of a previously short vowel, [u:], in all the Finno-Romani dialects with the partial exception of FIN-005 (where case number 132 “word” has to forms: [la:v] and [lau]). Case number 215 “outside” suggests that the original *v has stayed a [v] in RO-022 when it is not word-final (the RO-022 form is [vri]), but as there is only one case where this segment is not word-final and as Sinti is included in the analysis mainly in order to make the analysis of Finno- Romani clearer, this will not be discussed further. For this group I then reconstruct a *v. Whether the conditioning factor for *v to become [u] is the fact that *v occcurs in syllable- coda or if it is simply the fact that it occurs immediately after a vowel is not entirely sure. If the condition is only that *v occurs after a vowel, then this rule would predict that forms like 152 [ji:vela] or 186 [a:vel] would instead be [ji:uela] and [a:uel] respectively. This is obviously not the case. However, in both these two cases [v] is found after long vowels, therefore it is more likely that they stand in syllable-onset. This again points back to the condition which is concerned with *v's position in the syllable rather than its direct contact with a vowel. There are no cases where *v occurs in syllable-coda but not immediately after the vowel and indeed this might not be permitted by the phonotactics of Finno-Romani, in which case it is a somewhat moot point whether or not the rule ought to allow for this. Given the available data I prefer a rule that states that *v changes to [u] when it occurs in syllable-coda for the reason that it explains all the documented occurrences. Whether or not its claim is too broad must be reconsidered if any cases show up with the structure VCv in a single syllable. Until this happens, I note the following rule for Finno-Romani: 29
  • 30. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. *v → u / _ . (Complete with one partial exception: FIN-005: 132) [u2] Case number 10 “boy” appears to show a situation where either an original *u has become [o] in RO-022 making the the Sinti form [tʃao] (RO-022), or an original *o has become [u] in Finno- Romani making the Finno-Romani form [tʃau] (FIN-012). While this is not an impossible sound change, a comparison with all the other Romani dialects in the RMS-database and the European Romani forms strongly suggests that the original form of the word was actually [ʧavo]. This makes another course of events quite likely: First [o] disappeared in Finno-Romani (-o is a masculine suffix) and then the resulting [ʧav] underwent the rule stated above where a syllable-final [v] becomes [u]. RO-022 appears to have lost [v] instead, but I will not discuss this further. The [u]'s in this correspondence group then reflect *v rather than *u or *o. [e1+2] As for correspondences with [e], they seem to be quite stable generally, as there are no changes at the segment level across the dialects in the e-correspondences. I therefore reconstruct *e for both these correspondence groups without further discussion. [o1+2] This stability is also found in many cases with o-correspondences: [o] is found across all the dialects, and it seems reasonable to reconstruct *o for these cases. In the RMS-database case number 234 (“white”) is transcribed as having [ø] rather than [o] in FIN-012, but I have judged this to be a transcription error and have noted it as [o] instead. I have based this decision on the fact that 232 (“short”) too is noted as having [ø] in FIN-012, and in that word ([støt]) it is very clearly the case while in 234 (“white”) I for one cannot hear it. I thus do not note any changes for these groups. [o3] In two cases, 83 (“moon”) and 84 (“month”) FIN-002 has [u] where the other Finno-Romani dialects and Sinti have [o] (for instance FIN-002 [tʃu:n] and FIN-012 [tʃho:n]). A comparison with the other dialects in the RMS-database shows that these two cases are quite clearly reflexes of the same cognate even though they have different meanings. The two words also show up with identical forms in FIN-008, FIN-012 and AT-001x (and nearly identical forms in FIN-005, while FIN-006 has a gap and might also potentially have had an identical form). This is confirmed by Boretzky (2012: 42). Almost all other dialects where this form is found in either 83 (“moon”) or 84 (“month”) have 30
  • 31. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. [o] rather than [u]. The most obvious explanation of these facts is that [u] and [o] in 83 (“moon”) and 84 (“month”) are both to be regarded as reflecting *o and that something has possibly happened in FIN-002 to change this. As it is not possible to verify the transcription and as there are only these two cases which more or less only count for one, I will not attempt to explain this any further, and do not set up a rule for it. [i1+2] Many cases show [i] across all the dialects, and it seems obvious to reconstruct *i for these cases. [j1] Three cases (11 “girl”, 16 “horse”, 47 “mouth”) show a show a pattern where FIN-002, FIN- 005, FIN-006 and FIN-012 have [i] while FIN-008 and the two Sinti dialects have [j] – for instance is 47 [mui] in FIN-012 and [muj] in RO-022. In all three cases [i]/[j] appears right after a vowel and thus becomes the last part of a diphthong. A comparison with European Romani and all the other Romani dialects shows that the segment was originally [j] and these three cases then show a development parallel to [v] becoming [u] as shown earlier in this section: Certain consonants become parts of diphthongs in Finno-Romani when they appear in syllable coda. However, unlike the v → u change, this does not appear to happen in FIN-008 for some reason, although as there are only two cases to show that it does not and [i] and [j] can be hard to tell apart in syllable-codas, this might in fact be a general rule for Finno-Romani. Another explanation might be that the various transcribers of the dialects in the RMS-database have used slightly different transcription conventions. This does not seem particularly likely given the amount of work which has gone into unifying these conventions though (see Matras, White & Elšik 2009: 11), so I am simply noting that FIN-008 appears to behave differently in this regard: *j → i / _ . (Complete except in FIN-008) There are several correspondence sets where a vowel segment appears in one dialect but not in others. In order to decide whether these groups show sound addition or sound loss, I compare them to the forms listed by Boretzky and to the European Romani forms listed by Matras (2010). In cases where there was no European Romani form listed, I have compared the Finno-Romani forms with the other Romani dialects in the RMS-database and noted the most widespread form as an 31
  • 32. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. approximation of a protophoneme. [Ø1] In two cases a dialect has added [e] word-finally. In 6 “friend” it is FIN-006 ([ma:le]) and in 84 (“month”) it is FIN-005 ([ʦo:ne]). As it does not appear to be a general tendency in either of the dialects and as it happens in two different dialects but in both cases to nouns, it seems likely that this might be a morphological matter. There are several Romani nominal suffixes with the form -e but these are either plural markers or vocative suffixes (Matras 2002: 82-85) and it is not clear why the words in the given cases would display either outside of any context. However, as there are also no apparent reason for the variation in the segmental structure, it seems more likely that it should be a morphological matter and I will therefore not treat it any further. [Ø2] In 206 “now” Finno-Romani has lost a word-final [a] (also a word-initial one, but that development is shared by Sinti and is thus assumed to have happened before Northwestern split into a separate branch and is therefore not relevant here) which Sinti has kept ([ka:na]). As this is the only case where this has happened and as the word can be assumed to be quite frequent and therefore vulnerable to sound loss, I will not deal any further with this than to note that the Northwestern protoform of the word most likely has an *a. [Ø3] The European Romani form of 139 “pocket” reveals that the original form is kept in Sinti (The form [posita] which is also listed as common in the Northern dialects by Boretzky (2012: 71), while it has lost a syllable in Finno-Romani, giving it the form [posta]. 139 “pocket” is unusual in that the stress is on the second syllable instead of the first in RO-022, and this might have been the reason that this word alone has lost a word-medial [i]. The most likely course of change is that the stress moved to the first syllable and that the [i] was consequentially lost (syncope), but as there is only this one case, it is not possible to say for certain. It seems reasonable to construct *i for this segment, but as it appears to be a unique example, I will not attempt to state the rule governing this change. [Ø4] In 85 “wind” FIN-012 has acquired an extra [a] word-medially ([balava]). As with 139 “pocket” this seems to be a unique case, and I will not state any rules for it. Judging by European Romani and the other dialects in the RMS-database, this word did not have *a in the middle of the protoform, and according Granqvist (in press) mentions the case as an example of a quite recent 32
  • 33. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. process of epenthesis of a svarabakhti vowel in certain words in Finno-Romani. 5.2. Discussion of vowel inventories In summary of the analysis of correspondences with vowels, it seems, then, that on a segmental level, no great changes have occurred in the vowels of Finno-Romani since Proto-Northwestern. No segments have been lost and no new segments have been added. Apart from [v] becoming [u] and [j] becoming [i] post-vocally in Finno-Romani (which are not vowel changes in a strict sense) there are only minor, occasional differences between RO-022, AT-001x and the Finno-Romani dialects to be found, such as the vowel harmony happening in FIN-005 and FIN-012 and to some extent in FIN-008. Proto-Northwestern can then be assumed to have had a standard symmetrical five-vowel system with three heights: Figure 1 – The vowel system of Proto-Northwestern While the overall tendency in the cognates seems to be that the vowels have been quite stable in the Northwestern group of Romani, the picture is a little different when it comes to simple comparison between the various modern dialects and Proto-Northwestern. All the modern dialects still have all the vowel segments shown above, but when considering all the inherited words listed in Appendix 1 and not just the ones in Appendix 2, it becomes clear that all the dialects except FIN-002 also have [æ]. When including the borrowed words as well, FIN-002, FIN-006 and FIN-012 also have [y] and [ø], and FIN-002 also has [æ] although these are never found in the inherited words. FIN-005 lacks [y] but does have [ø], whereas FIN-008 seems to be closest to the vowel inventory of Proto- Northwestern as it has only added [æ]. Generally, then, although there are variations amongst the 33
  • 34. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. dialects, the modern dialects of Finno-Romani have gained three front vowels - [y], [ø] and [æ] - in addition to the original five vowels of Proto-Northwestern. This makes for a quite atypical vowel system which does not appear to be symmetrical: Figure 2 – Vowel inventory of most of the modern Finno-Romani dialects Although there have been relatively few actual changes to the vowels in Finno-Romani, the modern dialects thus do have vowel inventories which are quite different from that of Proto-Northwestern, although not all the segments in the inventories are found in the inherited material. The vowel inventory shown above in figure 2 appears to be identical to that of Standard Spoken Finnish (Suomi, Toivanen & Ylitalo 2008: 21). Of this inventory Suomi, Toivanen & Ylitalo (2008) note the following: That the mid-series /e/, /ø/ and /o/ are actually halfway between IPA [e] and [ɛ], [ø] and [œ] and [o] and [ɔ] respectively, that /a/ is about the same height as /æ/, and that the inventory is more symmetrical than it immediately appears: There are three height classes (/i/, /y/ and /u/; /e/, /ø/ and /o/; and /a/ and /æ/) and “no vowels not assignable phonetically to a class consisting of at least two vowels“ (Suomi, Toivanen & Ylitalo 2008: 22). Furthermore, the Finnish vowel system is described as being very stable and with barely any variation across dialects and registers (Suomi, Toivanen & Ylitalo 2008: 23). It would require a more specifically phonetic study and a finer transcription to verify if all these traits are shared by Finno-Romani, but at the phonological level, 34
  • 35. Astrid Monrad, 20071314. Thesis. Department of Linguistics, Aarhus University, summer 2012. these two languages do share the same vowel inventory. That is not to say that this is necessarily due to Finno-Romani being in contact with Finnish. (Valtonen 1968, citet via Granqvist 2002) suggests that contact with Hungarian has given Finno- Romani [y] and [ø], and that [æ] is a later loan from the Scandinavian languages. Granqvist does not comment on this (except to note that the sequence [æi] resembles old Vlax-dialects spoken in Russia, but he does not seem to be suggesting that there is any connection between the two (Granqvist 2002: 64)), but given the fact that [æ] is more widely found in the inherited lexicon than both [y] and [ø], it does not seem immediately obvious that [æ] should be the newest loan. This does not, of course, rule out the possibility that Valtonen is right at this point. It is also possible that [æ] simply is more easily incorporated in words with the phonology of Proto-Northwestern. As for [y] and [ø] coming from the Scandinavian languages (which I understand to mean the Germanic languages of Denmark and the actual Scandinavian peninsula), this is perfectly possible as they all do have the sounds in question or at least some that resemble them closely. There are also loanwords in Finno-Romani which are clearly of Scandinavian origin which do have some of these sounds (for instance 124 “bed” which in FIN-006 is [sæŋøs] although the same word has [e] and [o] instead of [æ] and [ø] in both FIN-002 and FIN-012), although I cannot find any Scandinavian words in the list containing [y]. While there thus might have been ample influence from the Scandinavian languages, I find it remarkable that the vowel inventories of Finnish and Finno-Romani resemble each other as closely as they do. Even is the [y], [æ] and [ø] originally entered Finno-Romani via the Scandinavian languages or Hungarian, the similarities suggest that Finnish has played and still plays a significant role in the shaping of Finno-Romani. Even if Finnish is not the source of the enrichment of the vowel inventory of Finno-Romani, it might be what prevents that the new segments are lost again. All of that being said, it is also clear that these changes involving the new vowels have only taken partial effect, as the inventory of the inherited material has only been affected to a small extent (as opposed to the rule turning syllable-final [v]'s to [u]'s which is found everywhere). Also, as was noted above, figure 2 is only true for the collected Finno-Romani dialects – not for each of the dialects on its own. I conclude, then, that having a given vowel appear in loanwords does not necessarily mean that it has also been adopted to be used in inherited words. This also explains why the reconstruction did not reveal any major changes: The sounds in borrowed words apparently do not automatically or directly spread to the inherited lexicon. In the case of [æ] though, it seems that a borrowed sound has spread. This might be a matter of it having been present in Finno-Romani for 35