“Language Change”
PROSODIC CHANGE IN PROGRESS:
EVIDENCE FROM ESTONIAN*
The problem I want to address today is prosodie change in progress. History is full of examples of prosodie change, but as a rule, change is noted when it has already taken place. I am fully aware of the controversy concerning gradualness vs. abruptness of linguistic change, and I am more than ordinarily conscious of the problems involved in trying to observe an ongoing process: more often than not, Heisenberg’s uncertainty principle applies, and the very fact of observation changes the process that is being observed. Nevertheless, I am going to attempt to describe what I consider to be prosodie change in progress; I hope to present a reasonable argument for the possibility of observing and documenting ongoing prosodie change. My test case will be the prosodie system of Estonian.
Estonian has received a certain amount of attention from linguists in the past twenty years, and I will not have to go into excessive detail in giving an overview of its prosodie structure. It is well known that Estonian has three degrees of phonemic quantity in vowels, and that it distinguishes between a single consonant and two kinds of geminates in intervocalic position. Usually this is referred to as three degrees of consonant quantity. This three-way quantity system interacts with a two-way opposition between “strong” and “weak” forms of word stems. There are numerous words in the language that maintain the same quantity throughout their respective paradigms, and they do not concern us at the moment. But a fairly large proportion of Estonian vocabulary is subject to what is called “gradation” or “grade alternation”, and what I have been calling “degree change”. This term refers to the phenomenon that a word stem may appear in two different phonetic shapes within the paradigm, traditionally called “weak” and “strong”. The conditioning factors for the occurrence of the two degrees were originally phonetic/phonological, and they can still be fairly well identified in Finnish; but in Estonian the conditioning factors have largely become opaque due to various sound changes as well as analogical restructuring, and the degree change has been morphologized. The appearance of the weak or strong degree in a particular form is now determined by inflectional class. For example, members of a large class of nouns have the weak degree in the nominative singular and partitive singular; the other twelve cases of the singular are in the strong degree. In the plural, on the other hand, nom. pl. and part. pi. are in the strong degree and the other twelve cases are in the weak degree in this class of nouns. There are other inflectional classes that have a different distribution of weak and strong degrees; the morphophonemics of Estonian is extremely complicated. The Dictionary of Correct Usage (Kull and Raiet 1976) gives 90 different inflectional classes for nouns/adjectives and 25 for verbs.
Word stems in Estonian were originally disyllabic, and the degree change affects the consonant on the border between the first and second syllable. Originally only plosive consonants were subject to degree change. The change was qualitative, when the intervocalic plosive was single, and quantitative, when the consonant was geminate. By qualitative change, we mean change of the plosive into a fricative or glide (or zero); by quantitative change, we mean the change of a long geminate into a short one or vice versa. Inflection of the word mõte ‘thought’ offers an example of a quantitative change:
An example of a qualitative change is provided by the word sõda ‘war’:
The original conditioning factor was the open or closed nature of the second syllable of the disyllabic sequence. The consonant between the first and second syllable of the stem was in the weak degree when the second syllable was closed, and in the strong degree when the second syllable was open. The syllable is open when it ends in a vowel, and closed when it ends in a consonant. The developments in the words ratas ‘wheel’ and rada ‘path’ illustrate the process:
The placement of the syllable boundary is automatic: the boundary occurs before a short intervocalic consonant, or within a geminate (or consonant cluster). Thus in *rattahan, the boundary is before /h/, the second syllable is open, and the stem appears in the strong degree. In ratas, the second syllable is closed, and the stem appears in the weak degree.
There have been some recent attempts (Prince 1980) to describe the prosodie structure of Estonian with reference to strong and weak syllables. I find this inadequate, for reasons that are too numerous to be reviewed at this moment; in the present context, though, I should like to point out that the process of degree change does not operate with syllables, but with disyllabic sequences. For purposes of degree change, consonants are to be defined as internal within a disyllabic sequence. It does not matter whether these consonants are syllable-initial or syllable-final; what does matter is that they are intervocalic within a disyllabic sequence. The placement of the syllable boundary is automatic and irrelevant from the point of view of the application of the process of degree change. Syllable boundaries get reassigned according to the general syllabication rule of Estonian, after the degree change has operated. For example, the word sõda ‘war’, nom. sg., is in the strong degree; the genitive sg. of the word, sõja, is in the weak degree (because the genitive originally ended in -n which has disappeared through a general sound change). The syllable boundary falls before the intervocalic consonant in both cases, and the first syllable is in the short quantity in both cases. In this type of word, it is the initial consonant of the second syllable that is subject to degree change. Prince (1980) claims that only the syllable rime matters for establishing whether the syllable is strong or weak, and that syllable-initial consonants are irrelevant; this is clearly not the case for determining whether a word stem appears in the weak or strong degree.
The two-way opposition between weak and strong degrees is older than the three-way quantity opposition. The three-way quantity opposition developed originally only in positions in which degree change also can occur, that is within a disyllabic sequence (but not at the boundary of two disyllabic sequences). The system arose in Estonian in conjunction with syncope and apocope that took place approximately 500 years ago. Kask (1972) puts apocope in the 13th century and syncope between the 14th and 16th centuries. Both processes affected short vowels of an open succeeding syllable after a long preceding syllable. Actually the two processes are two aspects of the same phenomenon, since both involved the loss of a short vowel after a long preceding syllable; the process is called apocope when the lost vowel occurred in word-final position, and syncope, when the lost vowel was word-internal. Syncope has not spread into all dialects to the same extent: the coastal dialect of the north-east is least affected, and it is this dialect that has not developed a three-way quantity system. The north-eastern dialect retains the old two-way system that Estonian used to share with Finnish, and that is still present in practically unchanged form in most Finnish dialects.
Analogical developments have led to the appearance of Q2 and Q3 also in stems that do not contain consonants liable to degree change; for example, there are long and overlong diphthongs in Estonian disyllabic words, e.g. laulu (gen.) - laulu (part.), with a long diphthong in the genitive and an overlong diphthong in the partitive.
This constitutes a bird’s eye view of the Estonian prosodie system, described at a very abstract level and omitting practically all phonetic detail. The problem is that the system does not appear to be completely stable. For more than a hundred years, linguists have felt that the scheme I have just presented does not do justice to the prosodie system. Two proposals have to be considered: first, the claim that there are really four quantities rather than three, and second, that the distinction between Q2 and Q3 is not a distinction of quantity, but rather a distinction of tone.
The question of the fourth quantity has been around for more than a century. Mihkel Veske, the first Estonian to receive a doctorate in linguistics (from the University of Leipzig in 1872), claimed to hear a distinctive difference in overlong intervocalic /k/’s in words of different morphological structure. There is an emphatic particle in Estonian, -ki, which can be added to any form of a noun or verb. Let us take an i-stem noun ending in /k/, e.g. tukk ‘firebrand’. Adding the emphatic particle to the nom. sg. of tukk, we get tukk + ki ‘even the firebrand’; the part. sing, of this word would be tukki. Veske said that he could hear a difference between the first tukki and the second tukki, and he called the first tukki the fourth degree of quantity.
Later scholars have likewise claimed that they can hear a difference between morphologically different words that would be homophonous if there were no fourth quantity. In particular, these forms involve pairs of partitive vs. short illative, and abessive vs. da-infinitive. The regular illative is formed with the suffix -sse, but there are some words—actually quite a large number—that also permit the formation of a short illative simply by providing the first syllable with overlength. For example, the word linn ‘town, city’ forms a regular illative, Unnas se, and a short illative, Unna. Originally it was just the intervocalic consonant that could be geminated, but at the present time the short illative can be formed from many other kinds of words. Take another example, a loanword kool ‘school’, with a long illative koolisse, and a short illative kooli. The words of which the short illative can be formed in addition to the morphologically regular illative have to be marked in the dictionary? there is no general rule as to whether the short illative is permitted or not. In case the short illative is used, the use of the long form can be restricted in various ways that are not relevant at the moment.
If speakers do indeed distinguish between the short illative and partitive in words like Unna and kooli, the illative would be, according to Polivanov (1928) and some younger scholars who have rediscovered him, in the fourth quantity, and the partitive would have regular third quantity. The other potential contrast between Q3 and quantity 4 is in sequences like võita abessive sg. ‘without butter’, and võita ‘to win’, da-infinitive of the verb võitma. (The abessive is presumably in Q4.)
I will return to the question of the fourth quantity after reviewing some of the claims that Estonian is really a tone language. There are two different problems involved here. One is the claim that the difference between Q2 and Q3 is really a tonal difference, so that Estonian has two degrees of contrastive quantity and that there are two different tonal patterns that provide a further distinction within the long degree of quantity. A variant of this is the theory of Harms (1978), according to which the differences between Q2 and Q3 are due to differences in stress accent rather than quantity. The other claim is that there are really four oppositions, and that the tonal distinction serves to separate the two kinds of overlength. This latter theory is a variation of the theory of four degrees of quantity: there are four oppositions, but the opposition between what some call Q3 and Q4 is tonal and not quantitative.
I have already referred to Polivanov, who made the first systematic attempt to treat Estonian as a tone language (1928). This treatment was popularized by Trubetzkoy. Polivanov claimed that different pitch contours distinguish between different cases in Estonian, to wit, between nominative, genitive, partitive, and illative. Here are some of Polivanov’s examples:
Unfortunately many of Polivanov’s examples are full of errors, not only in the phonetic descriptions of different word types, but also errors in morphology. For example, he forms the partitive of tuul ‘wind’ as tuule, and the partitive is really tuult. There is no free variation here; all words of this class take a dental suffix in the partitive. Polivanov, however, uses such nonexistent forms as members of his tonal minimal pairs.
The latest proponent of the tone theory is Helimski (1977), who claims to have found lexical tone in Estonian. Helimski’s claims are based on about a hundred words produced by one informant, without any attempt to make a recording of the pronunciations, not to mention any acoustic analysis or listening tests. I have not been able to discern any system in Helimski’s observations. He claims to have found tonal minimal pairs in the language, e.g., that the word maks ‘liver’ and maks ‘payment’ constitute a minimal pair, ‘liver’ having high tone and ‘payment’ having low tone. Helimski says that the informant agreed that there really is a certain distinction between the two words, but it remains unclear what the informant may have had in mind. Other minimal pairs offered by Helimski include koor ‘bark of a tree’ with low pitch and koor ‘choir’ with high pitch, and a few others. I am afraid that Helimski’s claims cannot be taken seriously.
Leaving the question of lexical tone aside for the moment, what most of these analyses have in common is a feeling, not always clearly articulated, that the three-quantity framework does not account for the way the language really functions. In my own opinion, the system is out of balance because there is a morphologically determined two-way opposition between weak and strong degrees and a phonological opposition between word-level durational patterns that sometimes parallels the morphological pattern, but more often does not. There are frequent contradictions, e.g., when the strong degree is in Q1, and the weak degree is in Q3. (The word nuga is an example, cf. below.) There are also many words that are not subject to degree change, even though their phonetic shape is not significantly different from those words that are (e.g., nuga alternates with a weak form, but vaga does not). And there are large numbers of homophonous forms of different morphological makeup (e.g., the nominative, genitive, and partitive of vaga all have the same form). Partial paradigms of the words Zopp ‘end’, hapu ‘sour’, nuga ‘knife’, and vaga ‘pious’ illustrate these points.
Different linguists see different ways in which the system might readjust itself, either through the development of an additional quantity opposition or the emergence of tone as a distinctive features. There is a third theory that I will add at this point, the theory of Mati Hint that the language is in the process of changing from a mora-counting prosodie structure to a syllable-counting structure. My own theory is that the prosodie system of Estonian is changing from primarily quantitative to basically accentual. By accentual I mean a phonological system in which stress, quantity, and tone all play a part. The way people—even linguists—feel about a language constitutes part of the data; if it is felt by a number of speakers, particularly by those who have a relatively highly developed consciousness about language, that the system is changing, one might just as well start looking for the cause of their intuition.
The system got out of balance some five hundred years ago with the development of the third quantity. At an earlier stage, there were just short and long syllables, no overlong ones; this earlier stage is reflected in folk songs, the meter of which operates with just short and long syllables. There are eight syllables in the folk song line; since the meter is syllable-counting, the folk song lines tend to preserve earlier forms—before the occurrence of syncope and apocope. It is necessary to take a closer look at overlength to understand some of the changes that have been taking place and are still in progress.
The phonetic characteristics of overlength include extra length on the overlong syllable itself, and a falling Fo contour on the overlong syllable. Both features can be explained as the result of a general tendency to “preserve the integrity of the word.” I have established—in other contexts, working mainly with English—that there exists a tendency to keep the duration of words approximately constant; thus a derived disyllabic word in English has approximately the same duration as the monosyllabic word that constitutes the base (e.g., speed is of the same duration as speedy, cf. Lehiste 1971). In Estonian syncope and apocope, the loss of a vowel after a long syllable appears to have been compensated for at the word level by the addition of extra length to that long syllable. All monosyllabic words in Estonian are overlong, since all monosyllabic words have resulted from the loss of a final vowel after a long first syllable; the first syllable of disyllabic words is overlong when the word was originally trisyllabic, with a short open second syllable, etc. I have proposed an explanation for the falling pitch contour found on overlong syllables (Lehiste 1978): the pitch contour is the result of transferring the pitch contour of the whole disyllabic word onto the overlong first syllable.
The tendency to maintain the durational integrity of the word has brought about another phonetic characteristic of Estonian disyllabic words, namely the inverse relationship between the durations of the first and the second syllable: a short first syllable is followed by a half-long second syllable, long first syllables are followed by somewhat shorter second syllables, and overlong first syllables are followed by the shortest second syllables.
Disyllabic words thus have three phonetic characteristics: length of the first syllable (which can be due to the length of the vowel of the first syllable, or to a combination of vowel and intervocalic consonant duration); Fo pattern applied to the word; and the ratio between the durations of the first and second syllable. I ran an experiment ten years ago (Lehiste 1975) with synthesized disyllabic Estonian words. In this set I used two minimal triples, taba-tapa-tappa and sada-saada!-saada. In taba-tapa-tappa, I changed the duration of the intervocalic consonant in systematic steps, and provided each synthesized word with a second syllable of three possible durations, characterizing the three quantities. The words were synthesized on a monotone. In sada-saada!-saada, I changed the duration of the first vowel in seven 20-msec steps and provided each word with three second syllable durations and three pitch patterns: monotone, step-down on the two syllables, and falling on the first syllable, with low second syllable. The total number of test items was 252. Listening tests were given in Tallinn to 26 reasonably monolingual subjects. Each listener gave four responses; thus there were 104 responses per stimulus for a total of 6552 responses for each set.
Tables 1 and 2 summarize the results. It is obvious that the pitch contour does not affect the perception of Ql very much, but that it has a rather dramatic effect on the perception of Q2 and Q3. The step-down pattern favors recognition of Q2, while the falling contour favors perception of Q3. The duration of the second syllable also has an effect. Here the perception of Q2 is not very strongly affected, but the shortest V2 favors judgments of Q3, and the longest V2 favors judgments of Ql.
TABLE 1.
Saada responses with different Fo contours (all V1 durations combined). V2 duration = 90 msec.
TABLE 2.
Saada responses with different V2 durations (all V1 durations and Fo contours combined).
On the basis of a statistical analysis of these results (Lehiste and Danforth 1977), I concluded that the distinction between Q1 and the other two quantities could be made on the basis of the first syllable duration alone; the distinction between Q2 and Q3 was strongly influenced by the pitch contour, and relatively less strongly determined by the duration of the second syllable.
These results leave no doubt of the perceptual significance of Fo. I find it also illuminating (and logical) that Fo should play a crucial role in distinguishing between the two “new” quantities. Duration separates short from long; this is the inherited opposition. Pitch contributes to the separation of long and overlong; this is the new Estonian development. It is important to recall here that there were some durations which could be changed from Q2 to Q3 by the pitch contour alone. In these intermediate durations, then, it was the pitch contour that the listeners used for making the distinction between the words. But pitch is not the only distinction; overlong Q3 is normally longer than long Q2 (recall the origin of Q3 as a case of compensatory lengthening).
The phonetic characteristics of overlength have been recently investigated by several Estonian linguists and phoneticians in connection with an apparent rediscovery of Polivanov’s ideas (Remmel 1975, Lippus and Remmel 1976, Eek 1977 and 1979). Particularly interesting in this connection is Remmel (1975). In analyzing 26 words produced by 14 subjects, Remmel found that overlong short illatives (as compared to overlong partitives) had a two-peaked pitch contour and that some test words showed a decrease in intensity in the middle of the overlong syllable, which Remmel equated with the Danish and Livonian stød. The stød-like effects were primarily found in diphthongs.
In a later experiment reported in Lippus and Remmel (1976), Remmel used a research technique in which listeners were interacting with a computer. Ten subjects were used who had been speakers in the previous study. Five of them had employed somewhat different pitch contours with the illative (as compared to the partitive); four had the stød. The reaction to computer-synthesized stimuli was obtained from three keys corresponding to the genitive, partitive, and illative cases, each key having ten positions. The duration of the test words was the same; what was changed was the pitch contour and the intensity of the signal during a 60-msec period containing the pitch peak. The results concerning the pitch contour were inconclusive, but an interesting correlation was found with regard to the stød-like feature. Auditors were asked to set the key for the proper value for the genitive, partitive, and illative for the word kaevu ‘well’. Genitive is in Q2, partitive is in Q3, and if there are four degrees of quantity, then the illative would be in Q4. The auditors set the value of an intensity drop at the peak of the pitch contour. For the genitive, the average setting was -3.6 db, for the partitive, the average was -8.2 db, and for the illative, the average was -6.8 db. Note that the drop for the illative was smaller than the drop for the partitive, while the opposite might have been expected. Remmel does not make any statement about whether the difference between a drop of 8.2 db and 6.8 db is significant or not. The results show nevertheless that a stød-like feature may be associated with overlength: the genitive, in Q2, had an average intensity drop of 3.6 db, while the two overlong cases had drops that were twice as large. However, this experiment provides no evidence for a possible distinction between Q3 and Q4.
Eek (1977) ran a perception experiment in order to determine whether the supposed phonetic differences between overlong partitives and illatives—such as difference in length, difference in pitch contour, and presence of something like stød—have perceptual significance. There were seven speakers, who read 23 sentences five times in succession. Thus there were 35 productions of each test item. The test words were embedded in sentences, and consisted of three words in the genitive, partitive, and short illative (pütt, tünn, kool), two words in the partitive and illative (maak, piim), and one pair of words for the distinction between an abessive and a da-infinitive (võita-võita). Four perception tests were run, consisting of different sets of the test items. I will review only the first test, in which 46 listeners were presented isolated words taken from the sentences and asked to identify the case of the word. There were 350 test words in all. Eek reports that 215 of them, amounting to 61%, were placed in the correct case category by more than half the auditors. This “more than half” sounds more impressive than it really is: the average identification for the partitive was 54%, and for the illative 46%. Eek did not report any significance calculations, but I am not ready to believe that either 54% or 46% are significantly different from 50%, i.e. from chance. A certain amount of error is always present in listening tests with human subjects; the correct score for the genitive was 93%, where 100% would have been expected if the listeners’ attention had not wandered during the listening task. This difference between 100% and 93% is just about comparable to the fluctuations around 50% (54% and 46%).
The 215 words (out of 350) that were placed in the correct case category by more than half of the listeners included all the words read in the genitive (94 items). That leaves 128 partitives and 128 illatives to be accounted for. Only 70 out of 128 partitives, and only 51 out of 128 illatives, were placed in the correct category by more than half of the listeners. It must be remembered that the listeners were faced with a forced-choice decision in each case. I interpret Eek’s results to mean that the listeners identified the genitive correctly and were simply guessing when they had to make a judgment regarding an overlong word. However, in 1977 Eek was unwilling to accept these results as conclusive counterevidence to the existence of four degrees of quantity, and assumed that the results reflected lack of accuracy in reading rather than absence of any distinction between partitive and illative.
By 1979 Eek seems to have changed his mind. In another paper published in 1979, he says that considering the so-called fourth quantity degree possible is the result of an overestimation of the phonetic data. Eek admits now that the majority of listeners do not distinguish Q3 and Q4; those who had tried to distinguish the partitive and illative had indicated as Q4 both the partitives and the illatives that had been read with the longest first syllable. These longer manifestations had generally been produced in sentence-final position; listeners called a sentence-initial illative partitive, and a sentence-final partitive illative. Eek now realizes that it was methodologically improper to ask listeners to compare words that had not been produced under identical conditions, and accepts the fact that listeners can hear phonologically irrelevant differences within a category.
Not everyone who likes the fourth quantity is ready to give it up, as Eek seems to have done. Viitso (1979), for example, takes the four quantities as proven and uses them to build a phonological system in which stress is eliminated as a phonological unit.
While the existence of an opposition between Q3 and Q4 can be dismissed on the basis of Eek’s listening tests, the studies just quoted have presented new information about the phonetic manifestation of overlength. The phonological status of overlength has received a new interpretation in some recent research by Mati Hint (1978, 1979, 1980).
Hint recalls that apocope occurred in Estonian in two word types: the final vowel was lost in disyllabic words like *jalka, which became jalg, and in trisyllabic words like *matala, which became madal. Hint reasons that this was really one and the same process : a short syllable had the value of one mora, while a long syllable had the value of two morae. He reformulates the rule, as it applied when apocope took place, to read that apocope happened when the final open syllable was preceded by at least two morae. Apocope did not apply when the final open syllable was preceded by only one mora, for example in words like kala. During the time of apocope, Estonian was a mora-counting language. The most important change between that time and the very recent past is the shift in the interpretation of what is considered long: now it is the over-long syllable that is considered equivalent to two morae, and the short and long degrees are counted as consisting of a single mora. Hint argues that contemporary standard Estonian grammar is based on opposing Ql and Q2 on the one hand to Q3 on the other hand. Take, for example, the placement of secondary stress. According to the standard grammar, in polysyllabic words with Ql and Q2, secondary stress falls on the third syllable, but in words with Q3, it falls on the second syllable. Hint formulates the rule for the standard language as follows: secondary stress must be preceded by at least two morae. (It is possible to establish the position of secondary stress in a word by objective measurements of duration.)
Standard grammar assigns words to inflectional classes on the basis of quantity. Stems in Ql and Q2 pattern in ways that differ from the patterns of stems in Q3. The types Hint concentrated on are trisyllabic in the genitive. The form of the partitive plural depends on the quantity of the first syllable: words in Q3 form their partitive plural with -i, while words in Ql and Q2 form the partitive plural with a diphthong and a dental suffix. Thus a word like endine forms a partitive plural endisi, while words like punane have a partitive plural punaseid. Note that endine patterns in the same way as the four-syllable word inimene, which has a partitive plural inimesiz the overlong first syllable equals two short syllables.
In these words, secondary stress should be falling on the second sellable after an overlong first syllable: éndìsi, töölìsi (this becomes even clearer in the comitative: éndìsega, töölìsega). Hint observes that current colloquial usage differs from that of the standard: he claims that such words are currently being stressed on the basis of syllable-counting rather than mora-counting, namely that the stress falls on the third syllable in words of four syllables, regardless of the quantity of the first syllable. The innovating pronunciation, according to Hint, is éndisèga rather than éndìsega. Thus there is a tendency to structure words into disyllabic sequences on the basis of a syllable count, rather than using the principle that an overlong syllable counts for two short syllables. The special nature of overlength is in the process of being eroded, and the new principle, which is taking over, does not consider overlength as a basis for the assignment of a word to a morphological class.
Hint’s specific examples were taken from the merger of the two types of words ending in -ne or -s which are trisyllabic in the genitive. The earlier pattern, codified in the standard grammar, required a distinction between them on the basis of the quantity of the first syllable: short and long yield a diphthong and a dental suffix in the part, pl.; overlong yields an -i without the dental suffix. Standard forms would be endisi, töölisi vs. punaseid, näljaseid; what has happened is the generalization of the form of punaseid to produce words like endiseid, tööliseid.
Hint’s arguments are based on a very extensive study of the realization of morphological patterns by Estonian schoolchildren. His tests consisted of 101 words, presented in various ways (e.g. in sentences where the blanks had to be filled with appropriate forms) to pupils in nine schools in different parts of the country. Public schools have eleven grades in Estonia; the tests were taken by pupils of the fifth, eighth, and eleventh grades. Over 800 pupils participated in the tests, and the number of forms produced by them was 20,700. The results show that the form with the diphthong is spreading at the expense of the form with -i. The spread is gradual in every respect: the innovation spreads gradually through the lexicon, the change is most extensive in the youngest group, and the change is spreading from the north to the south. Hint describes the change as a simplification of the rule for the formation of the partitive. The traditional rule requires specification of the number of syllables AND the quantity of the first syllable, whereas the new, expanding rule eliminates the specification of the quantity of the first syllable and requires only the specification of the number of syllables.
This is a highly simplified presentation of Hint’s monumental study. (Another aspect of his study involved presenting similar tests to large groups of Estonian teachers.) Hint considers a fair number of other morphological types and sees the same tendency in all of them. The pattern of morphological change constitutes evidence for ongoing phonological change; the increasing percentage of innovative forms in the usage of the youngest subjects makes it possible to predict the direction of the change.
This does not mean that Hint has abandoned the three-way quantity system for the description of present-day Estonian. It should be remembered that three quantities are needed to describe words that are not subject to degree alternations within a paradigm. Hint’s current point of view is in fact very similar to mine. Hint (personal communication, Sept. 1, 1980) believes that there are three contrastive quantities in stressed syllables. The first quantity is established on the basis of duration. The distinction between Q2 and Q3 may depend on tone in voiced syllables, and on duration in cases in which the contrastive segment is voiceless. Both Q2 and Q3 are long. In unstressed syllables, there are only two quantities: short and long. The unstressed long is neither Q2 nor Q3, but a neutralized long degree.
I would now like to put the development of tone as a distinctive prosodie feature in a larger framework. In 1978, I published a paper entitled “Polytonicity in the area surrounding the Baltic Sea.” In this paper, I reconsidered the situation around the Baltic Sea which Jakobson (1931) had used as an example for a Sprachbund based on features of tone. In general, Germanic and Baltic languages spoken around the Baltic Sea either have preserved Indo-European tone in some form (Lithuanian, Latvian) or have created new tonal systems (Swedish, Danish) . The relationship between Latvian and Livonian is of particular interest in the present context. The prosodie system of Latvian differs from that of Lithuanian in two basic respects: accent in Latvian is fixed on the first syllable in contrast to the free accent of Lithuanian, and Latvian has developed a third tone in addition to the two inherited Indo-European tones which it shares with Lithuanian. It is generally accepted that these two differences from Lithuanian are due to contact with Livo-nian, a Finno-Ugric language with stress on the first syllable.
The Latvian third tone, manifested as a glottal modification, is phonetically very similar to the Danish stød. It arose in Latvian in connection with the retraction of word stress to a first syllable that carried an original acute. (I am using the term acute to refer to the pitch pattern that appears in Lithuanian as a long falling tone and in Latvian as a long even tone.) In words that were already accented on the first syllable, the acute continued in Latvian as the long even tone. In words in which the word stress was retracted to an originally unstressed first syllable with the acute, the first syllable now carried the third tone, often referred to as broken tone, phonetically similar to the Danish stød. In classical three-accent areas, the sole historical source of the third tone is this reflex of Baltic and Slavic acute. In a number of other dialects, the broken tone goes back to both one of the reflexes of Baltic and Slavic acute and to all reflexes of Baltic and Slavic circumflex; thus the third tone may also appear in unaccented syllables such as affixes and endings.
Evidence for the claim that the development of the third tone is due to language contact is to be found in a closer study of Latvian dialects. The so-called Tamian (or Livonian) dialects of Latvian are spoken in areas for which there is historical and archeological evidence that they were formerly inhabited by Livonians. Most of the Livonians have gradually shifted to Latvian over the past 700 years. In some of the areas Livonian became extinct by the middle of the 19th century; it survives in Kurzeme, on the coast, in the speech of a few hundred Latvian-Livonian bilinguals. The Tamian dialects exhibit a considerable number of characteristics that are clearly Finno-Ugric in origin, and it is possible to trace the spread of many such characteristics into the standard Latvian language.
Although most Livonians have shifted to Latvian (and Livonian thus constitutes a substratum of Latvian, at least for the Tamian dialects), the language survives in Kurzeme and thus is available for investigation. It can be shown that during the centuries of adstratum relationship, Livonian has, in turn, been influenced by Latvian. The most dramatic parallel between Latvian and Livonian is the presence of tonal oppositions in Livonian. No other Finno-Ugric language has full-fledged phonemic tone. Livonian has been variously described as having three tones that are identical with those of Latvian (Posti 194 2); as having a phonemic opposition between presence and absence of stød (Zeps 1962) ; and as having an accentual system involving five types of stressed syllables (Viitso 1974). Although Posti (1942) argued that the rise of stød in Livonian is due to internal factors, I consider the theory of borrowing from Latvian (Décsy 1965) to be the more plausible one, especially in the light of extensive borrowing of many other features, both phonological and grammatical, from Latvian into Livonian. Taking all factors into consideration, it appears reasonable to assume that the development of tonal oppositions in Livonian is due to language contact, and thus can be attributed to the incorporation of Livonian into what Jakobson (1931) described as a Sprachbund of polytonicity around the Baltic Sea.
Jakobson included Estonian in the polytonicity-Sprachbund on the basis of Polivanov’s work. Even though I consider Polivanov’s descriptions inaccurate, I do not doubt that Estonian prosody includes a tonal component. At this point I should like to recall that Remmel found a reduction in intensity in overlong quantity which he identified with the Danish stød and the Livonian broken tone. I have suggested that the pitch contour characteristic of overlong quantity is due to the transfer of the pitch contour of a whole disyllabic sequence to the first syllable which became overlong in connection with apocope and syncope; this would support spontaneous tonogenesis in Estonian, without any necessary influence from neighboring languages. Nevertheless some areal factors may be noted that could conceivably argue for linguistic convergence. These include the presence of tone in Lithuanian and Latvian and its documented spread northward into Livonian— in the direction toward Estonia; the development of the long-overlong opposition in Estonian, with its associated pitch differences; and the most recent findings of stød-like phonetic features in overlong syllable nuclei. The presence of st0d in Latvian and Livonian and its embryonic emergence in Estonian are at least suggestive of linguistic convergence, even if they cannot be taken as conclusive proof.
Especially interesting in this connection is the recent evidence that some dialects of Finnish may likewise be in the process of developing contrastive tone. Niemi and Niemi (1980) carried out an experimental investigation of southwestern dialects of Finnish in which words of the type CVCV are developing a half-long vowel in the second syllable. Their subjects, too, were schoolchildren—14-year olds, 57 in all. As you may remember, the half-long vowel in such words is a well-established feature in Estonian. The appearance of a half-long vowel in the second syllable reduces the phonetic differences between words of the types CVCV and CVCW (e.g. sata ‘hundred’ - sataa ‘it rains’). One way to keep these words from becoming homo-phonous is gemination of the intervocalic consonant, and this is indeed observed in many instances. In the southwestern dialects, however, a difference in the Fo contour is likewise found on the second syllable: in CVCV, the Fo contour on the second syllable starts at a higher value than in CVCW, so that the subjective impression is of a rising tone in words like sata, even though the rise on the first syllable in sata is no different from that in sataa. In cases in which the difference in duration of the second syllable vowels has been neutralized and gemination of the intervocalic consonant does not develop, pitch may indeed acquire a distinctive function. It is not claimed by Niemi and Niemi that this development is due to language contact; it is nevertheless interesting to note that these dialects of Finnish are spoken in territories adjacent to the Baltic Sea.
Additional evidence for the existence of a polytonicity-Sprachbund is accumulating. For example, Ellen Niit (1980) studied the intonation of twelve sentences in the coastal districts of Estonia. There were 173 different subjects and a total of 2076 productions. The sentence intonation patterns recorded from these subjects were analyzed by means of a computer, and a number of parameters of the acoustical signal were correlated with each other. Niit found that the sentence intonation in Western Estonia, and especially on the islands, was chasracterized by a tonal contour in which the syllable following the stresed syllable had higher pitch than the stressed syllable. This feature is also found in Scandinavian accents. Niit appeared surprised that the Baltic Sea did not seem to function as a natural barrier. Of course there is abundant evidence that for seafaring people, seas facilitate communication rather than inhibiting it. An additional example is provided by a recent study by Vaba (1979) which documents extensive overlap in vocabulary between two peninsulas, one belonging to Latvia, the other to Estonia, facing each other across the Irben Strait within the Baltic Sea. The overlap in vocabulary is such that often it cannot be determined which language borrowed from the other. The only possible explanation is intensive contact across the sea.
The prosodie system of Estonian is changing within the larger framework of the Sprachbund of which it is a part. I doubt whether it is going to develop a fourth quantity; from a phonetician’s point of view, it appears more reasonable that if a distinction is needed, it will be based on a phonetic dimension that can be independently controlled. The most probable direction for development appears to be the emergence of an accentual system, in which stress, quantity, and tone all play a part.
I would also like to put the ongoing morphological changes that Hint has so extensively documented into a larger framework. The phonetic developments that resulted in the emergence of overlength also eliminated the phonetic features that conditioned the occurrence of weak and strong degrees of the same stem within a paradigm. This has led to the morphologization of the formerly phonetic alternations. Hint’s results show that the relative importance of the phonetic shape of the stem is gradually diminishing: the distinction between inflectional classes that was formerly based on the quantity of the first syllable of the stem is on the way out. I consider this to be a further step in the lessening of the role of phonetic factors in determining the shape of a paradigm; in other words, a step toward strengthening the relative autonomy of the paradigm as an element of linguistic structure.
The very special nature of the overlong quantity ultimately derives from the fact that it originally represented a disyllabic sequence. What we are observing at the moment is the accumulation of evidence that overlong syllables are losing the characteristics of disyllabicity. Metaphorically speaking, it seems as if the language is beginning to forget that overlong syllables were originally disyllabic sequences, and is readjusting itself to the original system: overlong syllables are beginning to be treated as if they were just single syllables, and polysyllabic words are again being structured into disyllabic sequences on the basis of syllable count. This is particularly clear in the readjustments in the placement of secondary stress, but becomes evident also from the changes in the assignment of paradigms to inflectional types.
Thus there are two changes at work at the same time: the change directed at eliminating the special status of overlength, and the change directed at reinterpreting the significance of the phonetic features accompanying over-length—from a manifestation primarily by quantity to a manifestation involving contrastive pitch.
One of my purposes in presenting this analysis was to show that it is indeed possible to study prosodie change in progress. If prosodie change could be studied only after it has already taken place, it would not be possible to make predictions. I am predicting now that Estonian will indeed develop into an accentual language. I hope that there will be a future generation of linguists interested in Estonian who will submit my prediction to a conclusive test, and I hope likewise that the language itself will survive long enough to make this test possible.
__________
* Paper presented at the Winter meeting of the Linguistic Society of America in San Antonio, Texas, on Dec. 29, 1980.
REFERENCES
Décsy, G. 1965. Einführung in die finnisch-ugrische Sprachwissenschaft. Wiesbaden: Harrassowitz.
Eek, A. 1977. Experiments on the perception of some word series in Estonian. Estonian Papers in Phonetics 1977 :7-33.
——. 1979. Further information on the perception of Estonian quantity. Estonian Papers in Phonetics 1979:31-57.
Harms, R. 1978. Some observations and hunches concerning Estonian prosody. Estonian Papers in Phonetics 1978: 31-4.
Helimski, E. 1977. Some preliminary data on lexical and tonal oppositions in Estonian. Estonian Papers in Phonetics 1977: 35-8.
Hint, M. 1978. Changes in the prosodical system of contemporary Estonian. Estonian Papers in Phonetics 1978: 39-43.
——. 1979. Minevikuline ja tulevikuline aines keelesüsteemis. Nórgaastmeline i-mitmus. Keel ja Kirjandus 22:142-49, 200-8.
——. 1980. Minevikuline ja tulevikuline aines keelsüsteemis. Prosoodiatüübi nihked ja selle tagajärjed. Keel ja Kirjandus 23:215-23, 270-8, 349-55.
Jakobson, R. 1931. Über die phonologischen Sprachbünde. TCLP 4. [Reprinted in his Selected Writings I: Phonological Studies, 137-43. ‘s-Gravenhage: Mouton (1962).]
Kask, A. 1972. Eesti keele ajalooline grammatika I. 2nd ed. Tartu: Tartu Riiklik Ülikool.
Kull, R., and E. Raiet. (eds.) 1976. õigekeelsus-sõnaraamat. Tallinn: Valgus.
Lehiste, I. 1971. Temporal organization of spoken language. Form and substance: Phonetic and linguistic papers presented to Eli Fisceher-Jørgensen, ed. by L. L. Hammerich et al., 155-169. Copenhagen: Akademisk Forlag.
——. 1975. Experiments with synthetic speech concerning quantity in Estonian. Congressus tertius internationalis Fenno-Ugris-tarum, 254-69. Tallinn: Valgus.
——. 1978. Polytonicity in the area surrounding the Baltic Sea. Nordic Prosody, 237-47. Travaux de l’Institut de Linguistique de Lund 13. Lund.
——, and D. G. Danforth. 1977. Foneettisten vihjeiden hierarkia viron kvantiteetin havaitsemisessa. Virittäjä 4/1977:404-11.
Lippus, U., and M. Remmel. 1976. Some contributions to the study of Estonian word intonation. Estonian Papers in Phonetics 1976: 37-65.
Niemi, J., and S. Niemi. 1980. Finnish half-long vowels, gemination and word-tone. Manuscript submitted for publication in Working Papers in Language and Language Use 1, Department of Languages, University of Joensuu.
Niit, E. 1980. The structure of the Baltic prosodie area and the place of Estonian dialects in it. Preprint KKI-17, Academy of Sciences of the Estonian SSR. Tallinn.
Polivanov, E. D. 1928. Vvedenie v jazykoznanie dlja vostokovednyx vuzov. Leningrad.
Prince, A. S. 1980. A metrical theory for Estonian quantity. Linguistic Inquiry 11:511-62.
Posti, L. 1942. Grundzüge der livischen Lautgeschichte. Mémoires de la Société Finno-Ougrienne LXXXV. Helsinki.
Remmel, M. 1975. The phonetic scope of Estonian: Some specifications. Preprint KKI-5, Academy of Sciences of the Estonian SSR. Tallinn.
Vaba, L. 1979. Ob estonsko-latyšskix jazykovyx kontaktax. Soviet Finno-Ugric Studies 15:71-5.
Viitso, T.-R. 1974. On the phonological role of stress, quantity, and stød in Livonian. Soviet Finno-Ugric Studies 10:159-70.
——. 1979. Problemy količestva v estonskom jazyke. Soviet Finno-Ugric Studies 15:1-17.
Zeps, V. J. 1962. Latvian and Finnic linguistic convergences. Uralic and Altaic Series, vol. 9. Bloomington: Indiana University Press.
We use cookies to analyze our traffic. Please decide if you are willing to accept cookies from our website. You can change this setting anytime in Privacy Settings.