Intonation of Israeli Hebrew

Intonation: Israeli Hebrew

I n t r o d u c t i o n

The term 'intonation' is used in its narrow sense to describe the variation of pitch by the change in the fundamental frequency of the voice (F 0 ) in the process of speaking and the overall pitch-form of the utterance. In its broader sense, which is used for this entry, the term also includes such phenomena as accent (above the level of the lexical stress) and its location, rhythm, duration, and intensity. However, the characteristics which are frequently attributed to the 'biological' rather than 'linguistic' code, such as general loudness (shouting vs. whispering), are beyond the scope of 'intonation', although the border between the 'biological' and the 'linguistic' in this domain is frequently blurred or disputed (Gussenhoven 2004).

G e n e r a l P r o s o d i c C h a r a ct e r i s t i c s o f I s r a e l i H e b r e w

The prosodic features of lexical stress in Israeli Hebrew have a lot in common with the characteristics of intonation as defined above, and therefore their behavior is strongly interrelated. Following the famous experiments by Fry (1958), which showed that the main properties of the stressed syllable in English are pitch height, vowel length, and (only finally) intensity, similar results were obtained for Israeli Hebrew (Laufer 1987:47). The special characteristics of pitch, namely, its deviation from general contour and non-abrupt falling tone of the final syllable (as opposed to the expected abrupt fall), etc., can be crucial for the perception of stress, but much research remains to be done in order to understand the nature of this phenomenon. It is very probable that central roles are played by vowel quality and other properties which were recently shown to be very significant for some Western-European languages, such as English and Dutch (Gussenhoven 2004:14).

Israeli Hebrew is not a tonal language, i.e., pitch contour is not used to distinguish minimal pairs of words lexically. Although the perceived difference between minimal pairs distinguished by stress (such as in ‫ד‬ ‫ﬠַ‬ ‫צַ‬ [ | tsa ͡ ad ] 'a step' vs. ‫ד‬ ‫ﬠַ‬ ‫צָ‬ [tsa͡ ±ad ] 'he marched') is, to a large extent, based on their pitch movement, it is the deviation from the general direction of pitch that is responsible for the perception of stress and not a specific and constant pitch contour.

Spontaneous speech is divided into Intonation Units-prosodic units which have "a coherent intonation contour" (Izre'el 2005 for Israeli Hebrew, following Du Bois et al. 1992, see part 3 for the detailed discussion). Usually, each unit has at least one accent-a stress on the level of the Intonation Unit (frequently these are called 'sentence-accents', as the common analysis is theoretically sentence-based, instead of being based on the prosodic units of the spoken language). Accents tend to occur on the lexically-stressed syllables. Syllables which do not carry an accent frequently have no perceived lexical stress either.

Pitch deviation from the main intonational contour is one of the principle clues for the location of accent. The best known type of such deviation is the so-called 'intonational peak', a local maximal height of pitch. If the accented syllable is not final or penultimate in the unit, Israeli Hebrew has a tendency to postpone this peak to the post-accented syllable, i.e., the pitch curve continuously rises through the accented syllable and reaches its peak in the beginning of the syllable which immediately follows the accent (as will be seen below in Fig. 4). However, this post-accented syllable is never perceived as prominent or accented and has no additional properties of accent besides the height of pitch, such as special length or a specific vowel quality. This feature has led © 2013 Koninklijke Brill NV ISBN 978-90-04-17642-3 to the analysis of Israeli Hebrew in terms of trochaic foot structure (Becker 2003), but it probably could also be analyzed as a late phonetic realization of the peak, called 'late peak' by Gussenhoven (2004:91). The only attempt at a comprehensive description of the intonation of Israeli Hebrew was made by Laufer (the data presented in this section are a summary of Laufer 1987; the references refer to this study if not stated otherwise). This research employed the approach of the British school of intonation analysis. This method considers intonational contour to be a sequence of sub-units that have their own contour, but no signifié of their own. Instead, they form a higher unit, which has a coherent and meaningful contour of pitch and is called by Laufer a group of words. The division into the sub-units is defined by the accents that appear on the important words.

The central and only obligatory sub-unit is the intonational nucleus, or, simply, the nucleus-the accented syllable of the last important word in the 'group of words' (or, simply, the last accented syllable in the group). It can be followed by a tail-all the unaccented post-nuclear syllables. The sub-unit, which can immediately precede the nucleus, is the head: it starts with the first accented syllable of the unit and ends with the syllable before the nucleus. Finally, the unaccented syllables before the head form the pre-head.

Laufer's study is characterized by an iconicemotional approach. Accents are claimed to mark the 'important words' iconically, while speaker's attitude towards the utterance and the addressee is considered to be the only signifié of the intonational pattern.

Laufer identifies six possible pitch contours of the nucleus-called nuclear tones-in Israeli Hebrew.

a. (light/small) fall-pitch descends from its average level to low level. b. deep fall-pitch descends from the high level to its low level.

c. rise-fall-pitch rises to a high level and then abruptly falls to low level. d. (light/small) rise-pitch rises from low level to average level or slightly above it. e. high rise-pitch rises from average level (or slightly below it) to high level. f. fall-rise-pitch falls from relatively high level to low level, and then rises to the average level.

Contours that end with a 'fall' descend to the low level in the nuclear syllable and stay low until the end of the unit. On the other hand, a 'rise' extends from the nucleus towards the final syllable (this feature is especially noticeable in the 'falling-rising' tone, which can actually only fall on the nuclear syllable and rise along the tail). Fig. 1 represents the nuclear tones according to Laufer's description. The head can have either a low or a non-low (usually high) tone and the same distinctive features characterize the pre-head. The high head can downstep towards the nucleus over a few intermediate levels or downdrift for emphasis by intermediate falls and smaller accented climbs. Analogically, the low head can emphasize the whole unit by upstep or updrift.

Out of the multiple possibilities of constructing intonational contours available-considering the existence of six nuclear tones, two non-obligatory head types, the same number of possible pre-heads and a non-obligatory tail-Laufer identifies nine possible patterns for Israeli Hebrew, as detailed below (parenthesis indicates optional sub-units). To be more precise, each pattern is actually a "family of contours" (56), as some of its sub-units may be expressed in various ways, and the final form can vary as a result.

(i). (low pre-head +) (low head +) small fall (+ tail) This pattern sounds 'complete and firm', a property which is attributed to all the patterns with a descending nucleus. It is used with any type of clause or its component part in order to convey "a cold, phlegmatic, estranged, restrained, uninterested emotionless attitude, disagreement, lack of excitement" or even "fury, threat, unfriendliness, unpoliteness, rudeness, arrogance" (72). © 2013 Koninklijke Brill NV ISBN 978-90-04-17642-3 (ii). (low pre-head +) high head + small fall (+ tail) or: high pre-head + small fall (+ tail) Laufer claims this pattern to be the most widespread in formal speech, such as news reading. Owing to its high (pre-)head, this pattern has more "power and vividness" in comparison to pattern (i) and the speaker sounds more "decisive, interested and heartful" (77).

(iii). (any kind of pre-head +) (high head +) deep fall (+ tail) This pattern with falling nucleus is used to express complete and firm utterances in a similar manner to the two mentioned above. However, the 'deep fall' imparts the impression of interest, excitement, friendly attitude and involvement of the speaker to the unit (81).

(iv). (low pre-head +) low head + deep fall (+ tail) This pattern shares its properties with pattern (iii). However, the 'low head' adds the impression of protest or criticism and in some cases signifies impatience (85).

(v). (low pre-head +) (low head +) + rise-fall (+ tail) This pattern also has the properties of the falling patterns. Affirmative statements marked in this way mean that the speaker is self-confident. Content questions asked by the falling-rising tone express protest; exclamatives give the impression of surprise (88).

(vi). (low pre-head +) (low head +) small rise (+ tail) In an affirmative sentence this pattern is used to maintain continuity, i.e., to mark that the speaker is about to say more, but with a simultaneous expression of coldness or protest. Most yes-no questions are asked with a rising nuclear tone, the small rise is used to express an indifferent question or protest.

(vii). (low pre-head +) high head + small rise (+ tail) or: high pre-head + small rise (+ tail) This pattern can mark an affirmative statement as 'colored' by encouragement or consolation. Otherwise, it resembles the previous pattern, but lacks the 'cold' approach. Content questions marked this way serve to create friendly relations with the addressee. Imperatives in this pattern sound like polite requests (97).

(viii). (low pre-head +) (high head +) high rise (+ tail) The 'high rise' serves for the expression of astonishment relating to the word that contains the nucleus. It is used for surprised yes-no questions and echo-questions (102).

(ix). (low pre-head +) (high head +) fall-rise (+ tail) This pattern can mark a stronger involvedness than the other rising patterns. An additional central function of this intonation is contrast with or restriction of the statement to a specific element, sometimes implicitly excluding other possibilities (106). Although Laufer's study was published in 1987 and many alternative approaches in general prosodic theory have since been proposed, there has been no other extensive research project dealing with the intonation of Israeli Hebrew. Extensive as it is, this study is not unproblematic and has been criticized both for the description of the data and the analysis.

The data presented in the research are partly based on artificial examples and non-natural speech (almost all the examples in the book are non-natural). In many cases, especially in the field of spoken language, this is a disadvantage. For example, the analysis of phrasal accents in spontaneous speech shows the tendency for a certain systematicity in its location. As Laufer takes the iconic approach, a large number of examples seem to represent controlled or rare patterns of accent, which affect the whole intonational contour. Kuzar (1989) made a partial attempt to implement Laufer's theory in authentic speech in his study of information structure in Israeli Hebrew. However, the focus of this implementation was not intonation itself, but rather the role that the partition of a clause into sub-units of an intonational pattern (location of head, nucleus, etc.) plays in its information structure ( Information Structure).

The central debate regarding theoretical models of intonational contours relates to assumptions regarding the number of existent nuclear tones. Since, as stated above, there has been no attempt to describe the intonation of Israeli Hebrew by means of any other prosodic theories, I leave aside the general discussion regarding the relative value of the contour approach as irrelevant for this entry. It is apparent in Laufer's analysis that every separate direction of non-complex contours (falls vs. rises), regardless of its 'size' (small vs. deep/ high), has its meaning on the level of the discourse dynamics: 'finality' for the falling tones and 'non-finality' for the rising. On the other hand, the size of the contour according to this approach corresponds to the emotional involvement of the speaker: 'coldness' for small contour, 'non-coldness' for a wider range, regardless of the direction of pitch. Therefore, it may be plausible to define these properties of contour separately as the basic elements in the system.

In addition, understanding of the prosodic features could probably be refined, if one adds the dimension of information structure to the iconic approach of the research. For example, most of the sentences used in the study to represent the contour of the 'deep fall' in pattern (iii) express great animation of the speaker. On the other hand, a close examination of the examples of pattern (iv) (low head + deep fall) reveals, that the deep falling nucleus occurs on a narrowly focused element, while the 'low head' marks the repeated-and therefore topical and textually given-items, which actually tend to be unaccented in the real speech. This adds complexity to the overall picture. What at first sight seems to be peculiar to the pattern turns out to be an independent property of its sub-units and the particular size of a contour can apparently be motivated not only emotionally, but also structurally.

C u r r e n t R e s e a r c h o n I n t on a t i o n i n I s r a e l i H e b r e w

Research on intonation in Israeli Hebrew since the work of Laufer is limited. It consists mainly of the research based on the Corpus of Spontaneous Israeli Hebrew (CoSIH), especially relating to the characteristics of Intonation Units, boundary tones, and the location of accent. Besides this, a small laboratory study was done by Mixdorff and Amir (2002) on the basis of eight recorded intonations of the same phrase.

As with any other spoken language, Israeli Hebrew speech is organized in segments characterized by their supra-segmental, i.e., prosodic, structure. These segments are called intonation units and have "a coherent intonation contour" (following Du Bois et al. 1992:17). The major cues for the recognition of intonation units in Hebrew have been found to be similar to those identified in many other languages. These are the rhythmic organization of the unit, i.e., initial rush and final lengthening, pitch reset and pause (Izre'el and Silber-Varod 2009). Among these, final lengthening is the highest-ranked factor, while pause is the lowest one and is frequently absent. Pitch reset has almost the same importance as final lengthening. Even in the case of penultimate stress of the last word in the unit, lengthening aligns with the final syllable and not with stress (Silber-Varod and Kessous 2008).

One or more intonation units form an utterance-an intonation unit that ends with a final non-continuing tone (fall or high rise), together with the possible preceding additional intonation units, which have a final tone that indicates continuity, as shown in fig. 2 (Izre'el 2009). The final terminating tone (" 'terminal break' . . . a falling tone by default" [Izre'el 2010:59]) is usually marked by the symbol ||, the continuing tone by | and the high rising tone by /. Mixdorff and Amir (2002) identify intonemes in Israeli Hebrew that are well-known crosslinguistically: "falls for declarative accents, rises in non-terminal position and very strong rises for question-final accents". A boundary tone can be observed, even if the last syllable is unaccented.

The framework of CoSIH makes use of a similar system of final tones (Izre'el 2010), but the high rising tone is analyzed as 'appeal' (following Du Bois et al. 1992) and the falling as the 'default' or 'unmarked' tone. The evidence for the last definition comes from its ability to occur in various discourse environments, when the categories usually marked by other tones are explicitly or pragmatically defined by other means. For example, content ('wh-') questions can terminate with the falling tone, as was already shown by Laufer (1987). In this case there can be no actual appeal involved, as in a self-question, but it can also be the case that the discourse relations of 'question' are established between the speakers by alternative means, such as extra-linguistic activity (e.g., passing the microphone in the course of an interview) or in an explicit lexical form (e.g., "I have a question"). Otherwise, their final tone is rising. Therefore, unlike other final tones, the falling tone marks nothing in addition to finality (see also Cohen 2009).

As for the rising tone, it can be concluded from the analysis of content questions above that its function is not to mark the question per se, but to establish 'question' relations between speakers. More precisely, it places the marked utterance for consideration by the addressee.

It must be noted that the high final rise as a marker of "proposing the unit for the consideration of an addressee" is not restricted to independent question clauses as it is in the written language. For example (Fig. 3) ‫היום‬ ‫שמעתי‬ ‫שערפאת‬ ‫טוב‬ ‫שמעתי‬ ‫שלא‬ ‫להיות‬ ‫יכול‬ ‫אוזן‬ ‫בחצי‬ ‫הזאת‬ ‫לקואליציה‬ ‫להצטרף‬ ‫הולך‬ ʃamati a-jom bexet͡ si 'OZEN | jaxol (li)jot ʃ-lo šamati TOV | ʃe-ARA'FAT olex leit͡ staref la-koalit͡ sija a-zot / 'I heard today with half an ear, probably I didn't hear properly, that Arafat is going to join this coalition?' (capitals mark accented words, apostrophes mark stress-bearing syllables and not glottal stops) The unit that is presented for the consideration of other speakers in this case is the complement clause of the verb ‫שמעתי‬ ʃamati 'I heard'. Generally, the rise in yes/no questions falls on the questioned element. The combination of the high accent and the rise results in a very high tone, which keeps its pitch level until the end of the unit. In content ('wh-') questions the rise happens on the final syllable of the utterance-which is frequently unaccented (the main accent in content questions usually takes place on the initial indefinite (question) word). As a result, the actual rise reaches only the average pitch level. Therefore the final tone of yes/ no questions is significantly higher if compared to the rise of the content questions.

Imperatives frequently terminate with a rise, probably with a similar function of appeal to the addressee (for a similar cross-linguistic observation see Hirst and Di Cristo 1998:26). It must be noted, however, that these units end significantly lower than the high rises. In such cases imperatives generally also tend to be marked by a higher pitch on the accented verb itself (or immediately following it in accordance with the 'late peak' principle above) in addition to the small rise at the end of the intonation unit. This is the case in fig. 4, which shows the utterance ‫הביתה‬ ‫שלנו‬ ‫הבן‬ ‫את‬ ‫תשלחי‬ tiʃle±xi t-a-±ben ʃlanu a±bajta / 'Send our son home' with a significant ('late') peak, which terminates immediately after the accented syllable of the verb, in addition to the final rise.

The "rises in non-terminal position" (Mixdorff and Amir 2002), which can be seen in the first two units of fig. 2, mean that additional units follow within the same utterance and that the speaker keeps the turn to himself. In Fig. 2. The utterance xaʃuv ʃe-u javin | ʃe-me-oto rega ʃe-u alax | a-kvut͡ sa niret tov joter || 'It is important that he understands, that since the moment he left, the group looks better' is divided into three intonation units, the first two end with a continuous tone (|), whereas the falling final tone of the last unit (||) indicates finality (from Izre'el 2009) xa∑uv ∑eu ja v i n ∑emeotorega ∑eua l a x akvu t͡ sa nir e t t ov joter xaʃuv ʃeu ja v i n ʃemeotoroga ʃeua © 2013 Koninklijke Brill NV ISBN 978-90-04-17642-3 these units pitch tends to rise until the end of the unit, even in the case of penultimate accent on the last word, i.e., through the last syllable despite the lack of accent on it (Silber-Varod and Kessous 2008). In addition, Silber-Varod (2010) identified a boundary tone that marks hesitation in speech termed continuous elongation. Its most prominent cue is the rhythmic aspect-a significantly long final vowel, which can be an appended [AE] or a prolongation of the final syllable. Its pitch is a constant level-tone in the average key of the speaker. The most frequent occurrence of this boundary tone is attested in monosyllabic function words, such as ‫ה-‬ ha-'the', ‫ו-‬ ve-'and', ‫זה‬ ze 'this', etc.

An additional prosodic feature of Israeli Hebrew that received some attention is the distribution of accents in intonation units (Ozerov 2010). In prosodic theory there is a longstanding debate of this issue regarding the relative merit of the iconic and the structural approaches. According to the former approach (presented in Laufer 1987), the accent is a wholly iconic feature-the element, which is considered to be important by the speaker, receives the accent. The most distinguished representative of this school is Bolinger (e.g., 1972), who adduced a wide range of exam-ples to support his position. The structural approach claims that the accent obeys welldefined rules, based on syntactic criteria and/or information structure. This approach is found in Gussenhoven (1983), Lambrecht (1994), Selkirk (1995) and many others, usually based on controlled recordings.

Analysis of spontaneous speech in Israeli Hebrew suggests that the middle path is the most appropriate analysis: language tends to follow structural rules in the marking of accent. This pattern is the most widely attested and the most usual in everyday conversation, but not obligatory, since 'iconic' accents can be added under various conditions.

The approach presented in Lambrecht (1994) best fits the analysis of the data found in Israeli Hebrew. According to it, every element in a phrase has (1) its pragmatic role-whether it functions in the domain of a 'topic' or of a 'focus' (= 'rheme'/'comment', using a different terminology); and (2) its pragmatic featureswhether it was mentioned in the immediate previous discourse ('given') or is in the speaker's attention at the moment of the speech act ( Information Structure).

The element which is both 'given' and already functions as a topic in the discourse or has its 'prototypic' properties, e.g., a nominative (1) the nominative pronoun ‫זה‬ ze 'this' serves as a topic and carries no accent.

On the other hand, non-subject pronouns are not 'prototypical topics', and therefore their pragmatic-'topical'-relation to the proposition must be established by an accent, such as the same pronoun ‫זה‬ ze 'this' in (2). It is the final part in (2) that functions as the new information, while ‫זה‬ ze 'this' refers to the established topic of this discussion. Despite this, the non-subject position of the 'topic' results in an accent.

Members of the limited category of nouns which can occupy the subject position when appearing for the first time in the discourse (i.e., as 'non-given' information) also receive an accent in this function, as their pragmatic feature of 'non-given' is not suitable for an established topic. A topic can also be separated into its own intonation unit as the result of its functions in the discourse (Netz 2003).

Unlike 'topic', the 'focal' ('rhematic'/ 'comment') domain must have at least one accented element, i.e., the 'focal relations' are never established in advance. The nominative pronoun ‫הוא‬ hu 'he' in (3) is accented, as it is the only element of the 'focal' domain.

The accenting rules in this domain tend to follow the syntactic structure. The verb forms a single domain with its complement argument, while the accent usually culminates on the argument. If any non-accented lexeme intervenes in this domain, it creates a prosodic 'valley' and results in the perception of accents on both the verb and the argument, despite the fact that their prosodic properties are similar to those in the case of the single non-divided domain.

In nominal phrases only the last element is accented. In (4) the accent reveals that the verb is complemented by a single nominal phrase, ‫טיפול‬ ‫לאוטו‬ tipul la-'OTO 'car service (lit. service to-the-car)'. The alternative possible analysis of two verbal complements-direct object ‫טיפול‬ tipul 'service' and indirect object ‫לאוטו‬ la-'oto 'to-the-car'-would require a separate accent on each noun, and this is not the case here. Instead, in accordance with the rules of a single nominal phrase, the accent appears only on the final lexeme, ‫אוטו‬ 'OTO 'car'.

Pronominal arguments represent a given topical constituent if complement a verb and are not contrasted. Non-argumental adverbials that pertain to the 'focal' domain usually carry a separate accent.

From a cross-linguistic point of view, the accent rules show a certain degree of similarity to the rules of European languages, such as English or Dutch (Gussenhoven 1983:28). The question as to whether there is a special accentual pattern for an 'all-new' clause in Israeli Hebrew requires additional research. It is very probable that the accents have an additional role at a higher prosodic level and that the properties of the accent mark the role of the intonation unit within the utterance.

In content ('wh-') questions the question word is accented, although some question words, such as ‫איזה‬ ±eze 'which' are proclitic, as they form a part of a noun phrase. A stronger ('contrastive') accent on the question word marks an exclamation, usually with an opposite meaning ('Why did you do this' means 'You should not have done it'). However, this meaning does not exclude the possibility of the high rising terminal tone, which presents the unit for the consideration of the addressee and requests his or her opinion and explanation (i.e., the 'question' and the 'exclamation' do not represent mutually exclusive sentence types, but rather compatible linguistic categories).

From a phonetic point of view, Mixdorff and Amir (2002) found that the narrow focus accent "not only increases the duration of the syllable, but also stretches the following syllable". The duration of the syllable in the narrowly focused element is dependent on its usual or expected length: the ratio of the lengthening is significantly greater for usually unaccented pronouns than for verbal or nominal elements. As in European languages, such as German, no additional accents can follow the narrow focus until the end of the unit. On the other hand, unlike German, additional accents can precede the narrowly focal element. The nature and the behavior of the contrastive accent in spontaneous speech are as yet poorly understood.

C o n c l u s i o n

The intonation of Israeli Hebrew has been partially described, but has yet to be fully investigated. Its general characteristics, such as final and nuclear pitch contour, location and nature of the accent, and the principles of the division into intonation units, resemble many features of some of the better described European languages, such as English, German, or Dutch. A closer analysis, however, reveals some intriguing and significant peculiarities. In general, the few sub-fields that have been described and analyzed so far provide a wide scope for further investigation and research.