This preview has intentionally blurred parts. Sign up to view the full document

View Full Document

Unformatted Document Excerpt

BeUer 3 Phonology For or For Worse W1U-1.tJJ A::AO AN'1ELLMe.IF ? by Lynn Johnston , FOR BETIER OR FOR WORSE 1990 Lynn Johnston Productions. Dist. By Universal Press Syndicate. Reprinted with permission. All rights reserved. 3 .0 What Is Phonology? oth phonetics and phonology can be generally described as the study of speech sounds, but they are not the same field. Phonetics (the subject of Chapter 2) is specifically the study of how speech sounds are produced, what their physical properties are, and how they are interpreted. Phonology, on the other hand, is the study of the distribution of sounds in a language and the interactions between those different sounds. Phonologists ask the following kinds of questions: What is the organization of sounds in a given language? Of all the sounds in a language, which are predictable and which are unprediCtable in given contexts? Which sounds affect the identities of words? B Contents 3.1 The Value of Sounds: Phonemes and Allophones Introduces thl? two levels ofphonological representation-phonemes and allophones-and describes the three basic ways sounds can be distributed in a language. 3.2 PhonolRgic;I1 Rules Describes how phonological rules map between the two levels, introduces the idea of natural classes, and. i1Jtroduces several types of common phonological processes. 3.3 Phonotacticc::onstraints and Foreign Accents Introduces the idea that there are language-specific limitations on how sounds can be put together, and relates this to some of the reasons that non-native speakers of a language seem to have a foreign accent. 3.4 Implicational Laws Describes how certain phonological patterns recur in languages, in a particular ordered and introduces some explanatory principles for these patterns. 3.5 How to Solve Phonology Problems Outlines some basic techniques and strategies for solving phonological analysis problems. 3.6 Practice Provides exercises, discussion questions, and activities related to phonology. 100 3.1 The Value of Sounds: Phonemes and Allophones 3.1.1 Predicting the Occurrence of Sounds In both Kikamba (a Bantu language spoken in Kenya) and English, we can hear the sounds [k] and [g]. The Kikamba word [kosuuIJga] 'to guard' contains both phones, as does the English word [kagnelt] cognate. The difference between Kikamba and English lies in the way the two sounds contribute to the identity of a word. In English, the two phones can distingUish words, as shown by words like [trek] tack and [treg] tag! where alternating between [k] and [g] affects the message conveyed by the utterance. IIi this sense, phonologists say that the occurrence of these two sounds in English is unpredictable, since we cannot look at the rest of the word and determine which sound will occur. That is, if we know that a word in English begins with [tre] , we cannot predict whether the word will end with [k] or [g] since both tack and tag are different, but possible, words. In Kikamba, on the other hand, the sounds [k] and [g] are predictable from their environment. Sounds are predictable when we expect to see one sound but not the other based upon the sounds that precede and/or follow it. In Kikamba, the only consonant that can come directly after an [IJ] is [g], and [g] can only come immediately after [IJ]. The combination [IJk] does not occur in Kikamba (see Roberts-Kohno 2000). So, if there is a velar stop in a word in Kikamba, we can predict whether it will be a [k] or a [g]: it willbea [g] if it is immediately preceded by [IJ]; otherwise, it will be a [k]. However, . inpnglish we cannot make this prediction: the sound [k] does appear after the sound [IJ], as in [reIJkJ,] anchor! as does the sound [g], as in [reIJgJ,] anger. To illustrate how strong this distribution is in Kikamba, consider the case where you have a word with a [k] such as katala! the base form from which conditional forms of the verb 'to count' are built. To say 'if you count,' you add an [0] to the front of the word: [okatala]. But to say 'if I count,' you add an [IJ]. Even though this word has a [k] in it, we have seen that it is a rule in this language that [k] cannot appear after [IJ]--so the [k]appears as a [g] instead: [IJgatala] 'if I count.' This type of alternation does not happen in English, because there is no rule governing the distribution of [k] and [g]. So while Kikamba and English both use the phones [k] and [g], the languages differ in that in Kikamba we can predict the occurrence of one versus the other,and in English we cannot. If someone learning Kikamba were to use [k]after [IJJ, the identityofthe word would not change. Instead, a native speaker of Kikamba might think that the speaker sounded funny, had an accent, or had mispronounced the word. On the other hand, if a learner of English were to make the same substitution in English, then the identity of the word is likely to change. Imagine confusing [k] and [9] and saying at a dinner party "we're haVing [glreb] for dinner tonight./1 Your guests might feel rather uncomfortable, especially if they were expecting crab! The bottom line is that in Kikamba, the sounds [k] and [g] are predictably distributed, while in English they are not. 101 102 3.1.2 Allophones and Phonemes In every language, certain sounds pattern together as if they were simply variants of the "same" sound, instead of different sounds that can be used to distinguish words, even though they may be phonetically distinct. For example, the sounds [k] and [9] are clearly different sounds: we use them to make the contrast between different words in English, as we saw in Section 3.1.1. But as we also saw, these two sounds are completely predictable in Kikamba. In Kikamba, then, these sounds can be thought of as variants of the "same" sound, because in any given context, if there is some velar stop sound, we can predict which one ([k] or [9]) it will be. Similarly, if you ask a native speaker of English how many different sounds are represented by the underlined letters in the words !lin, !lin, and s!lin, they will probably say "two," grouping the aspirated [ph] of pin and unaspirated [p] of spin together. Though [ph] and [p] are phonetically different sounds, native English speakers often overlook this difference and may even consider them to be the "same" sound (see Section 2.6.5). One of the goals of this file is to help you understand more clearly the distinction between "same" and "different" sounds. To do this, we will discuss the terms allophone and phoneme. Since these concepts are the crux of phonological analysis, it is important that they be clearly understood. Perhaps the best way to start to explain these terms is through examples. On a separate piece of paper, transcribe the following five words in IPA: (1) top stop little kitten hunter It is likely that you transcribed all of these words with a [t], like the following: (2) [tap] [stop] [htO [hAnq:] This is good, since it probably reflects something that is psychologically real to you. But, in fact, the physical reality (the acoustic phonetic fact) is that the 't' you transcribed in those five examples is pronounced slightly differently from one example to the next. To illustrate this, pronounce the five words again. Concentrate on what the 't' sounds like in each example, but be sure to say them as you normally would if you were talking to a friend-that is, don't try to enunciate them abnormally clearly. What differences did you notice? Compare, for example, the ItI of top to that of stop. You should be able to detect a short burst or puff of air after the It I in top that is absent in stop. That puff of air is what we have called aspiration (see Section 2.6.5), which is transcribed with a superscripted [h]. So while a native speaker might think of the 't' sound in top and stop as being the same sound, the 't' is actually pronounced differently in each word. This difference can be captured in the transcription, as in [thop] and [stop], respectively. Now say the words little and kitten. We might say that the 't' in little sounds" softer" than the one in stop, and is clearly voiced. For most speakers of American English (but not British English), the 't' in words like little is pronounced as a flap, [f], much like the rin Spanish in words like [pafa] 'for' and [tOfO] 'bull' (see Section 2.2.5). English kitten, on the other hand, is pronounced with the same sound we hear in the expression uh-oh, a glottal stop [1]. So, we could transcribe little and kitten $is [lIfll and [kI11,l], respectively. For some speakers of American English, in casual speech words like hunter are pronounced with no 't' at all, but rather as [hAnf]. Try to say it this way and see if it sounds like something you've heard before. In any case, while you may have initially transcribed the five words above with a Itl, they may also be transcribed in a way that reflects the different pronunciations of that sound, as in the following: (3) [thop] [stop] [hfll File 3.1 The Value of Sounds: 103 To a native speaker, all of the words above Seem to have a 't' in them, at least at some psychological level. 1 Evidence of this lies in fact that one may transcribe them all with a 't,' at least until trained in transcription. Someone who lacks linguistic training would probably not hesitate to state that all the words have a it' and would need to be convinced that subtle differences, like exist among them. In this sense, the above words do have a 't.' On the other observe that the It' may be pronounced in several different ways. Unlike a speaker of English, a native spe'ak,:f Hindi does not the difference between aspirated and unaspirated sounds when spl::aldng or hearing Hindi. To a speaker of Hindi, the aspirated sound [ph] is as different [pJ as is from [b] to our ears. The difference between aspirated and stops must be noticed by Hindi speakers because their language contains many that are pronounced in nearly the same way, except that one word will have an stop where the other has an unaspirated stop. The data in (4) illustrate this. (4) Hindi Gloss [phal] 'fruit' 'moment' 'strength' [pal] [ball A native speaker of English may not be aware of the difference between aspirated and unaspirated stops because aspiration will never make a difference in the meanings of English words. If we hear someone say [mcep] and [mceph], we may recognize them as different pronunciations of the same word map, but not as different words. Because of the different ways in which [pI and [ph] affect meaning distinctions in English and Hindi, these sounds have different values in the phonological of the two languages. We say that these two sounds are noncontrastive in English, since two does not result in a change of m.eaning. In Hindi, on the other contrastive, since replacing One sound with the other in a word can the We will have more to say about this terminological distinction below. Linguists attempt to characterize these different relations sounds in language by grouping the sounds in a language's sound into classes. Each class contains all of the sounds that a native speaker considers as "same" sound. For example, [t] and [t h} in English would be members of the same class. But [t h] and [d] are members of different classes because they are contrastive. one for the other in a word, you can cause a change in the word's [thmm] time versus [dmm] dime. On the other hand, speakers of Hindi would not as m.embers of the same class because they perceive them as different. That are contrastive in Hindi. A class of speech sounds that seem to variants of the same sourid is called a phoneme. Each member of a particular class is caUed an allophone,which corresponds to an actual phonetic segment produced That the various ways that a phoneme is pronounced are called aWJpllOfles. In this view, we can say that the It' SOlmc!s like stop, top, little, and kitten all belong to a single class, which we will label by the <;vrnh"l characterizing this particular phoneme. By saying that stop and top, for eX;;imjole, each have the phoneme /t/, we are saying that the sounds It] and [t h] are related. I IThe reasons for this may be manifold, including ph,om:tic similarities, ph,ow)!o,gical patterning, difvarieties, or spelling. ferent pronunciations across 104 In (5) we see how the phoneme It I is related to its allophones in English and how the Hindi phonemes It I and Ithl are related to their allophones. In English, [t], [t h], [r], and [?] are allophones of the same phoneme, which we can label It/. In this way, we can say that in English the phoneme It I has the allophones [t] as in [stop], [t h] as in [thop], [r] as in [h4], and [1] as in [krfl(l]. On the other hand, in Hindi, [t] and [th] are allophones of different phonemes. Note that symbols representing phonemes are written between slashes; this distinguishes them from symbols representing (allo)phones, which are written between square brackets. (5) English Phonemes: Allophones: H indi It I [t] [?] [r] By proViding a description like this, linguists attempt to show that the phonological system of a language has two levels. The more concrete level involves the physical reality of phonetic segments, the allophones, whereas phonemes are something more abstract. In fact, linguists sometimes describe phonemes as the form in which we store sounds in our minds. So, phonemes are abstract psychological concepts, and they are not directly observable in a stream of speech; only the allophones of a phoneme are. The phoneme is a unit of linguistic structure that is just as significant to the native speaker as the word or the sentence. Native speakers reveal their knowledge of phonemes in a number of ways. When an English speaker makes a slip of the tongue and says [tfi kem] for key chain, reversing [tf] and [k], he or she has demonstrated that [tf] functions mentally as a single unit of sound, just as [k] does. Recall from File 2.2 that [tf] is phonetically complex, consisting of [t] followed immediately by (fl. Yet, since [tf] represents the pronunciation of a single phoneme Itfl in English, no native speaker would make an error that would involve splitting up its phonetic components; you will never hear [ti kfen] as a slip of the tongue (see File 9.3). Knowledge of phonemes is also revealed in alphabetic spelling systems (see File 13.4). For example, English does not have separate letters for [ph] and [p]; they are both spelled with the letter p. Examples like this show that the English spelling system ignores differences in pronunciation that don't result in meaning distinctions. For the most part, the English spelling system attempts to proVide symbols for phonemes, not phonetic segments. In general, alphabetic writing systems tend to be phonemic rather than phonetic, though they achieve this goal with varying degrees of success. As noted in File 2.1, of course, there are multiple ways to represent the same sound (e.g., the [k] sound is written with a <k> in the word kitten but with a <c> in the word cool). What's crucial here, though, is that both of these spellings represent Ikl, and not, say, the difference between [k] and [kh]. 3.1.3 Identifying Phonemes and Allophones: The Distribution of Speech Sounds In order to determine whether sounds in a given language are allophones of a single phoneme or allophones of separate phonemes, we need to consider the distribution of the sounds involved. The distribution of a phone is the set of phonetic environments in which it occurs. For example, nasalized vowels in English occur only in the environment of a nasal consonant. More precisely, a lingUist would describe the distribution of English [I], [6], and so on, by stating that the nasalized vowels always and only occur immediately preceding a File 3.1 The Value of Sounds: Phonemes and 105 nasal consonant. In this book we will mainly be concerned with two types of distributioncontrastive distribution and complementary distribution-though a third distribution, free variation, will also be introduced in the following section. Let us consider contrastive distribution first. Recall from above that a pair of phones is contrastive if interchanging the two can change the meaning of a word. This means that the sounds can occur in the same phonetic environment. It also means that the sounds are allophones of different phonemes. Two sounds are noncontrastive if replacing one phone a nother does not result in a change of meaning. Our earlier discussion of the patterning of [pJ and [PhJ in Hindi and English provides a good example of this difference. Recall that we said that in Hindi these two sounds could affect the meaning of a word based on examples like [pdlJ moment and [phdlJ fruit, where the tWQ meanings are distinguished by the o<;:currence of [pJ or [PhJ. This means that the two sounds [pJ and[phJ are contrastive in Hindi. In English, on the other hand, simply replacing [pJ for [phJ, or vice versa, will never effect a change in the meaning of a word; the sounds are noncontrastive in English. We just determined whether or not [pJ or [phJ are contrastive in Hindi and English by taking into account the distribution of sounds in each individual language. We did this by identifying a minimal pair. A minimal pair is defined as a pair of words whose pronunciations differ by exactly one sound and that have different meanings. When you find a minimal pair, you know that the two sounds that differ are contrastive and, thus, the sounds involved are allophones of different phonemes. If you try, you can think of many minimal pairs in English, or any other language you know well. For example, the minimal pair [thi:mJ team and [thi:nJ teen shows that [nJ and [mJ are allophones of separate phonemes (that is, they are contrastive) in English since they can be used to contrast meaning. In Hindi, the words [phdlJ 'fruit' and .[bdlJ 'strength' constitute a minimal pair, showing [phJ and [bJ to be allophones of separate phonemes; [phdlJ fruit and [pdlJ moment also form a minimal pair in Hindi. But notice that there are no minimal pairs involving [ph] and [pJ in English; these two sounds are never contrastive with respectto one another. Instead, they are allophones of the same phoneme, /p/. Consider another example in which two languages make different distinctions using the same set of sounds. In English, [IJ and [lJ are contrastive phonemes, as can be seen from the existence of minimal pairs such as leaf [lif] versus reef [lifJ, alive [dlmvJ versus arrive [dlQIV], or feel [filJ versus fear [filJ. In Korean, on the other hand, [IJ and [rJ are never contrastive. 2 Consider the data in (6). (6) Korean [IJ versus [rJ alternations Citation Form Nominative Case Gloss [pulJ [malJ [talJ [khalJ [pudJ [madJ [tadJ [khadJ [palJ [salJ [padJ [sadJ 'fire' 'language, speech' 'moon' 'knife' 'foot' 'flesh' 2you will notice that the two "r" sounds in English and Korean are not the same phonetically-in English, it is a voiced alveolar retroflex liqUid [I], while in Korean, it is a voiced alveolar flap [r] (see Section 2.2.5). Similarly, the exact articulation of /11 in the two languages is also not identical, though we use the same symbol for both. These phonetic differences are not particularly important here, however, because we are concerned only with the distribution of the two sounds in each language (a phonological question) rather than with the quality of the two sounds (a phonetic question). 1 06 Notice that in each example, only one English translation is given. For example, the word for 'fire' has two forms: one when it is used by itself, the citation form (such as in the answer to the question "What is the Korean word for 'fire'?"); and one when it is used as the subject of a sentence, the nominative form (such as in "The fire burned brightly"). This use of different forms for different grammatical roles is similar to the use in English of, say, atom as the noun for the smallest unit of an element ("An atom of helium has 2 protons") as opposed to atomic as the adjectival form of the same word ("The atomic structure of helium is simple"). In the English case, it's clear thatthe ending -ic at the end of atomic is a marker of an adjective. Similarly, in Korean, the final -rin the second column of words in (6) indicates that these words are in the nominative case. (For more on these kinds of markers, see Chapter 4.) If we remove the special ending of the word atomic, you can. see that the base of the word is still atom. But notice that the base atom in the word atomic is not pronounced the same as it is in isolation. Theword atom by itself is pronounced ['a:,ffP.], but in atomic, it's pronounced [.d'thom]. Given that these two forms represent the same base word, we might expect that they would be pronounced the same way. Instead, the pronunciation of the base word alternates depending on the phonetic context it appears in. An alternation is simply a difference between two (or more) phonetic forms that you might otherwise expect to be related. Identifying alternations relies on the assumption that, all else being equal, the same word should be expressed by the same we find different pronunciations ofthe same word that are systematically linked to particular grammatical contexts, we have an alternation. So in English, the base word atom is expressed alternately by the sounds ra:,flll] and [.d'tham]. In Korean, the base word fire is expressed alternately by the sounds [pull and [pur]. Why might these words have alternating pronunciation.s depending on what other elements they appear with? The answer to this question takes us back to looking at the distribution of the sounds [1] and [f] in Korean. In the words listed in (6), we see that [1] and [f] do not occur in the same phonetic environment. Even though we are looking at the "same" word (for example, the word for 'fire'), the phonetic quality of the third sound is not always the same-it alternates between [1] and [fl. Specifically, as you can see from the data in (6), you use an [1] when the sound is the last sound in the word (is in "word-final" position), but you use [f] when the sound is between two vowels (is in "intervocalic" position). In fact, if youwere to look at all of Korean, you would find that [f] appears only between two vowels, while [1] never appears in that position. Meanwhile, [1] can appear at the ends of words, but [f] never does. 3 These two observations mean that if someone gave you the frame [to_] in Korean, you could tell them whether an [f] or an [1] goes in the blank. In Korean, it must be an [1]1 Notice that you cannot do the same thing in English: either an [1] or an [:r] could go in the blank to form a possible English word. The difference that has been illustrated here between English and Korean is that in English, [1] and [1] are in contrastive distribution, while in Korean, [1] and [f] are in complementary distribution. Sounds showing this type of distribution are considered to be allophones of the same phoneme. To understand better what we mean by complementary distribution, think about what the term complementary means: two complementary parts of something make up a whole. For example, the set of people in your class at any given moment can be divided into the set of people who are under 5'5" tall and the set of people who are 5'5" tall or taller. These two sets of people complement each other. They are mutually exclusive (one person can't simultaneously be both shorter and taller than 5'5''), but together they make up the whole class. The Korean sounds [1] and [f] are in complementary distribution, because they appear in different sets of environments: [f] occurs between vowels, and [1] occurs word-finally. 3In some modern words that have come into Korean from other languages, [1] can also appear at the beginning of words. But it never appears as the only segment between two vowels. File 3.1 The Value of Sounds: Phonemes and 107 Given our assumptions about alternations-that the same word should be expressed by the same sounds-we can hypothesize that even though (I] and (r] are phonetically different phones in Korean, they are allophones of a single phoneme. We can represent that single phoneme as Ill. At this point, it probably seems like an arbitrary choice as to why III should be the phoneme and not hi; we will talk more about how to make this choice in File 3.5. If two sounds are in complementary distribution in a language, there will never be a minimal pair that uses them to distinguish two words. Furthermore, the appearance of one allophone or the other will always be predictable, as we saw above with the frame (ta_]. You can predict that (r] but not (1] will appear between vowels in any word in Korean, and that [l]but not (r] will appear word-finally-even if you have never studied Korean. This kind of prediction is a powerful tool in helping phonologists understand the structure of languages. What's particularly interesting about this (and all other) phonological distributions is that it represents actual knowledge tnat native speakers have. For example, if you give a native speaker of Korean the new (nonsense) word moladam and ask them to say it out loud, they will say it with an (r] between the two vowels, and not an (I]! Of course, this is not something that anyone has explicitly taught them (especially since they have never seen this word before), but the distribution of sounds is one of the things that you know when you know a language (see File 1.2). Consider another linguistic example, namely, the distribution of the English sounds [p] and [Ph] shown in (7). (7) spat spool speak [speet] [spul] [spik] pat pool peek [pheet] [phul] [phik] As you can see in the English words in (7), [p] and [ph] do not occur in the same phonetic environment. As a result, there are no minimal pairs involving a [P]_[ph] contrast. In fact, the phones are in complementary distiibution: [p] occurs after [s] but never wordinitially, while [ph] occurs word-initially but never after [s]. Since these sounds appear in different phonetic environments, there can be no pair of words composed of identical strings of sounds except that one has [p] and the other has [ph]. As stated above, phones that are in complementary distribution are allophones of a single phoneme. In .this case, [p] and [ph] are both allophones of the phoneme we can represent as Ip/. Furthermore, the appearance of one allophone or another in a given context is predictable. For example, we can predict that the allophone [Ph] (but never [p]) will appear in word-initial position. So even in words not listed in (7), we know that it will be [ph], rather than [p], that will occur at the beginning of a word. 4 Similarly, we can predict that [p] (but never [ph]) will follow [s] in other words. 3.1.4 Free Variation Most phonological distributions can be described as either contrastive or complementary. Remember that the hallmark of a contrastive distribution is that you can't predict which of two (or more) sounds belongs in a certain context, because each will produce a different but meaningful word; the hallmark of a complementary distribution is that you can predict which of two sounds belongs in any given context. In some contexts, however, more than one pronunciation of a given sound may be possible. In these cases, you may not be able to predict exactly which sound will occur, but the choice does not affect the meaning of the 41n point of fact, this is true not just at the beginning of a word but at the beginning of any stressed syllable. That is, in English, [Ph] but not [pI can appear as the first consonant of a stressed syllable. 108 word. Consider, for example, the pronundations of some English words in (8) (remember that [17'] represents an unreleased voiceless bilabial stop). (8) leap soap troop happy [lip] [soup] [trup] [ha:pi] leap soap troop [lip'] [soup'] [truP'] *[ha:p'i] These words show that [17] and [17'] both share some of the same phonetic environments; spedtically, they can both appear at the ends of words. Unlike the case of English [b] versus [ph], or [m] versus [n], however, there are no minimal pairs involving these sounds in the language. Why not? Although there are pairs of words in (8) that differ in only one sound, none of these words contrast in meaning. Thus, the choice between [17] and [17'] in leap/ soap/ and troop does not make a difference in meaning; that is, the sounds are noncontrastive. Rather, they are interchangeable in word-tinal pOSition. Sounds with this type of patterning are considered to be in free variation. To a native speaker, sounds like [17] and [17'] that are in free variation are perceived as being the "same" sound. We can conclude that they are allophones of the same phoneme, because they are perceived as the same and do not serve to distinguish the meanings of words. Another term that you may encounter to describe the different types of phonological distributions in language is overlapping. If two sounds are in overlapping distribution, they can occur in the same environment. Both sounds that are in contrastive distribution and sounds that are in free variation are therefore considered to have an overlapping distribution; only sounds that are in complementary distribution do not overlap. For example, in English, the sounds [d] and [1] are in overlapping distribution because they can contrast: the words lid and lit form a minimal pair, and both [d] and [t] can occur after [h]-that is, the environment [bJ is one where [d] and [t] overlap. Similarly, [t] and [t'] have an overlapping distribution because they can also both occur after [1IJ, as two different pronunciations of the word lit. The difference between [d] and [t] on the one hand, and [t] and [C] on the other, is. that interchanging [d] and [t] changes the identity of the words, while inter[t'] does not. 3.1.5 Summary To summarize, a phone's distribution is the collection of phonetic environments in which the phone may appear; when linguists describe a phone's distribution, they describe this collection. Relative to each other, two (or more) phones will be in contrastive distribution, in complementary distribution, or in free variation. Phones in contrastive distribution may appear in minimal pairs and are allophones of different phonemes. Phones in free variation may, like phones in contrastive distribution, also appear in the same phonetic environments/ but never cause a contrast in meaning; they are allophones of the same phoneme. In either of these two types of distribution, given a particular phonetic enVironment, one canw hich of the phones will occur. If the phones are in complementary distribution, their appearance in particular phonetic environments is predictable; they never appear in minimal pairs, and they are allophones of the same phoneme. 3 .2 Phonological Rules 3.2.1 Phonological Rules In File 3.1, we discussed the fact that phonemes and (allo)phones belong to different levels of structure in language-that is, phonemes are abstract mental entities, and phones are physical events. In this file we consider the connection between these two levels. The mapping between phonemic and phonetic elements is accomplished using phonological rules (recall from Section 1.2.3 that a rule of grammar expresses a pattern in a language). A speaker's knowledge of phonological rules allows him or her to "translate" phonemes into actual speech sounds; knowledge of these rules formspartofthe speaker's linguistic competence. This change from the phonemic underlying form to the actual phonetic form of a word by means of phonological rules can be represented with a diagram: (1) phonemic form V- rules V- phonetic form As an example, consider the English word can Ika:n/. This word has a finallnl sound in its phonemic form, and in fact it is frequently pronounced with a final [n]. If we listen carefully, however, we find that the final consonant of can (especially in casual speech) is often [m] or [1)].1 The examples in (2) illustrate this. (Here and throughout this file we use a fairly broad transcription style, recording phonetic detail only for the segments under discussion.) (2) I can ask I can see I can bake I can play I can go I can come [01 kren resk] [01 kren si] [ai krem beIk] [01 krem pleI] [01 kre1) gou] [01 kre1) kAm] (or [01 kIf resk]) (or [01 kIf si]) (or [01 kIp. beIk]) (or [01 krp pleI]) (or [01 kl) gou]) (or [m kl) kAm]) As these transcriptions show, the phoneme Inl is pronounced as the phone [m] when it precedes a labial consonant and as the phone [1)] when it precedes a velar consonant. We can state this fact about English as a descriptive rule: lIn linguistic analysis, we often have to distinguish between "careful" and "casual" speech. Careful speech is when a speaker speaks more slowly and dearly than usual, while casual speech is when a speaker speaks more qUickly and with more co-articulation (see Section 2.2.6) than usual. Of course, these are really endpoints on a scale of speech styles (see File 10.1), and people actually talk at many different styles in between. 109 110 (3) Inl is pronounced as [m] before a labial consonant [lJ] before a velar consonant [n] everywhere else. (We will be adjusting this rule later on in this file.) Notice that a phonological rule has three parts: the sound(s) affected by the rule, the environment where the rule applies, and the result of the rule. Here Inl is affected by the rule. The rule applies when Inl is followed by a labial or velar consonant. The result of the application of the rule is that Inl acquires the same place of articulation as the following consonant. We can write this rule using shorthand of the form X Y I C _ D. Here, 'X' is the sound that is affected by the rule, 'Y' is the result of the application, and 'c _ D' is the environment in which the rule applies (the conditioning environment). By "C _ D," we mean that C comes before the sound in question, and D comes after it; the blank represents where the sound appears. You can read these rules in the following way: "X becomes Y when it comes after C and before D." Thus, if you saw the form CXD, you would know that it turns into CYD if the rule applies. So, for the rule in (3) above, we would write: (4) [n] [n] [n] [m] I_labial consonant [lJ] 1_ velar consonant [n] I everywhere else Now consider how the phonetic forms of some of the above examples are derived from the phonemic forms: (5) phonemic form: apply rule: phonetic form: Ikren reskl k<en resk [kren resk] Ikren belkl krem belk [krem belk] Ikren goul krelJ gou [krelJ gou] This illustrates what happens in speaking. In listening, a hearer reverses this process: he or she perceives the phonetic form of an utterance, then sends it "backwards" through the phonological rules, and finally obtains a phonemic form that matches a form stored in memory. The rule illustrated above applies not only to In I, but also to It I and Id/: (6) hat trick hit batsman night class bad dream head band bad guy [hret tlIk] [hlp bretsmq.] [nmk klres] [bred dlim] [hEb brend] [breg gal] 3.2.2 Natural Classes Can we make one rule to state that Inl, Itl, and Idl all change place of articulation according to what sound follows? Is it random chance that these three sounds all seem to undergo the same phonological rule? To answer these questions, let's first take a look at the articulatory descriptions of these three sounds: (7) It I Idl Inl voiceless alveolar oral stop voiced alveolar oral stop voiced alveolar nasal stop Not only are all three sounds alveolar stops, but they are the only alveolar stops in English. (Note that there is no such thing as a voiceless alveolar nasal stop as a phoneme of Rules F ile 3.2 III English.) Therefore, we can make the description more general by removing some of the properties: (8) In I, Itl, Idl alveolar stop With respect to English, saying "alveolar stop" is the same as saying Inl, Itl, and Id/. These three sounds are all of the phonemes in English that are produced by stopping the flow of air at the alveolar ridge. Thus, they are the natural class of alveolar stops. A natural class is a group of sounds in a language that share one or more articulatory or auditory property, to the exclusion of all other sounds in that language. That is, in order for a group of sounds to be a natural class, it must include all of the sounds that share a particular property or set of properties, and not include any sounds that don't. All of the properties used in Files 2.2 and 2.4 to describe individual sounds can also be used to describe natural classes. For example, in the English vowels the monophthongs [i, u] and the first part of the diphthongs [eI] and foul are tense vowels, and there are no other tense vowels in English. Thus, these vowels are members of the natural class of tense vowels in English. Likewise, the consonants [k, g, lJ] are all described as velar consonants, and they are the only velar consonants used in English; thus they constitute the natural class of velar consonants in English. Notice that we already referred to the natural class of velar consonants in the formulation of our rule at the beginning of this file. You'll recall that this rule affects the natural class of alveolar stops when followed by a member of either the natural class of velar consonants or the natural class of bilabial consonants. This shows that natural classes can be used to describe both the sounds affected by a rule and the environments where a rule applies. In talking about groups of sounds, we must use a few properties in addition to those needed to describe individual sounds. For example, if you look at the consonant chart on the last page of this book, you will notice that the only labiodental consonants in English are the fricatives [f] and [v], while the bilabial fricative slots are left I n many situations it is advantageous to refer to [f] and [v] together with [p, b, m, w] and ['Y'] as belonging to the same natural class. For this purpose we use the property labial. Another property used to describe natural classes divides the segments into two groups, obstruents and sonorants. Obstruents are produced with an obstruction of the airflow. The sounds in this category are stops, fricatives, and affricates. Sonorants, on the other hand, are segments produced with a relatively open passage for the airflow. Sonarant segments include nasals, liquids, glides, and vowels. Thus, the class of labial obstruents in English is [p, f, b, v], while the class of labial sonarant consonants is [m, w, 'Y' J.Z The class of labial consonants is the union of both sets: [p, f, b, v, m, w, 'Y'}. As we will see, being able to divide consonants into obstruents and sonorants is qUite useful in stating phonological rules. 3.2.3 Classification of Phonological Rules In addition to seeing that phonologlcalrulesapply to naturaldasses.of segments, we can classify phonological rules according to the kind of process that they involve. Seven major kinds of processes are discussed here, along with examples from the phonology of English and other languages. a. similation. Rules of assimilation cause a sound (or gesture) to become more like a neighboring sound (or gesture) with respect to some. phonetic property. In other 2As already mentioned, the clgSs of sonorants also includes vowels, because they do not have an oblirlguists treat rounded vowels . " in which case the entire class strlletion of English labial sonorants would be [m, w, 'Ytu, U 0, J], and the labia Is would be [p, f, b, v, m, w, 'Yt ti, U, o,J]. f 112 words, the segment affected by the rule assimilates or takes on a property from a nearby (often adjacent) segment. Rules of assimilation are very common in languages. The first rule we considered in Section 3.2.1 falls into this category. We can call it alveolar stop assimilation because it applies to all alveolar stops (ltl, Idl, and In/): (9) Alveolar stop assimilation (English): Alveolar stops assimilate to the place of articulation of a following consonant. Thus, when a sound haVing the properties alveolar and stop immediately precedes a labial consonant, this rule causes the alveolar stop to take on the property labial (thereby replacing its specification for alveolar). Similarly, this rule can apply to change the sound's place of articulation feature to dental when it precedes a dental consonant (examples such as width [WIc!e] and in this [II;:). OIS]), and so on, for the other places of articulation. We saw examples of this sort of alveolar assimilation in (2) and (6). These examples of assimilation all took place across a word boundary (e.g., the Inl in can assimilating to the Ibl in bake). We can see a similar sort of phenomenon taking place across word boundaries in certain ASL handshapes. We will consider the handshape that is used in the sign ME, which is a pointing index finger, as shown in (10). (10) The unassimilated sign for ME in ASL 1j ''''''''''L.'' V ( /? I r\ W U J 2006, William Vicars, www.Lifeprint.com. Used with permission. The sign ME may take on features of other handshapes, however, depending on the sign that follows it. For example, in order to say "I am named ... ," a speaker of ASL would sign "ME NAME .... " In order to say "I know," a speaker of ASL would sign "ME KNOW." The signs NAME and KNOW have different handshapes: NAME is articulated with two fingers (index finger and middle finger) extended; KNOW is articulated with a bent hand and all four fingers extended. When the sign ME is produced before one of these other words, it can take on the handshape of the word that follows it, as shown in (11). (11) a. The phrase 'I am named. ,.r.' in ASL, formed from the lexical items ME NAME 2006, William Vicars, www.Lifeprint.com. Adapted by permission. File 3.2 Rules 113 Notice that in both (lIa) and (lIb), the signer touches his chest with his hand facing the same way as he would in the unassimilated form. That is, the place of articulation, the orientation, and the movement for ME do not change. But the handshape used for ME in (1Ia) is the handshape of NAME, and the handshape used in (lIb) is the handshape of KNOW. Another assimilation process is palatalization. Palatalization refers to a special type of assimilation in which a consonant becomes like a neighboring palatal. For example, when American English speakers say Did you? rapidly, they very often pronounce it as [dId3U]. The sounds [d] (the alveolar stop from the end of did) and [j] (the palatal glide from the beginning of you) combine to form the palatal affricate [d3]. In this case, the palatal nature of the glide has been assimilated by the stop, making it a palatal affricate. Front vowels such as [i] and [e] also cause this change. The most common types of palatalization occur when alveolar, dental, and velar stops or fricatives appear before a front vowel. So the following are all common types of palatalization: [t] rtf]; [d] [d3]; [J]; [k] rtf]; [g]H [d3]. While there are variants on palatalization, and other sounds can be palatalized, the main things to look for are a sound becoming a palatal and/or a sound change conditioned by a front vowel. The rules of assimilation that we've discussed so far cause sounds to assimilate to adjacent sounds. This is a common way that assimilation occurs. However, long-distance assimilation also exists, and a relatively common type of long-distance assimilation is called vowel harmony. This typically causes all the vowels in a word to "harmonize" or agree in some property such as. rounding or backness. Finnish has a common type of vowel harmony rule, which can be stated as follows: (12) Vowel harmony (Finnish): A back vowel becomes front when preceded by a front vowel in the same word. By this rule, Finnish words have, with few exceptions, either all front vowels or all back vowels, but not both in the same word. We can see the vowel harmony rule in action when a suffix is added to the end of a word. In this case, the suffix vowel changes to match the quality of vowels in the word. For example, the suffix meaning 'in' has the form ["sso] when added to a word where the last vowel is back, as in [tolo] 'house,' [tolosso] 'in the house.' However, the suffix takes the form [-ss<e] when it attaches to a word with a final front vowel, as in [mets<e] 'forest,' [mets<ess<e] 'in the forest.' In cases like this, we can say that the vowel of the suffix harmonizes, or assimilates, to the preceding vowel. b. Dissimilation. Unlike assimilation, which makes sounds more similar, rules of dissimilation cause two close or adjacent sounds to become less alike with respect to some property, by means of a change in one or both sounds. An example of dissimilation in Greek is the follOWing: (13) Manner dissimilation (Greek): A stop becomes a fricative when followed by another stop. For example, in fast speech especially, the form lepta! 'seven' can be pronounced as [efta], and /ktizma/ 'building' can be pronounced as [xtizma] ([x] is a voiceless velar fricative). c. Insertion. Phonological rules of insertion cause a segment not present at the phonemic level to be added to the phonetic form of a word. An example of this kind of rule from English is voiceless stop insertion: (14) Voiceless stop insertion (English): Between a nasal consonant and (lvoiceless fricative, a voiceless stop with the same place of articulation as the nasal is inserted. Thus, fOf instance, the voiceless stop insertion rule may apply to the words dance /d<ens/ [d<ents], strength /stJ IJEl/ [stlH]kElJ, and hamster /ha:mst.j:/ [ha:mpst.!:]. 114 d. Deletion. Deletion rules eliminate a sound that was present at the phonemic level. Such rules apply more frequently to unstressed syllables and in casual speech. English examples include: (15) Ih/-Deletion (English): /h/ may be deleted in unstressed syllables. The /h/-deletion rule would apply to a sentence such as He handed her his hat /hi heendad h+hIZ heet/ to yield [hi heendad +IZ heet). Deletion is common in fast speech because it saves time and articulatory effort. Sounds like [h) that are not very perceptible are often the "victims" of deletion because speakers can save time and effort by deleting them without sacrificing much information. That is, the listener may not be relying on these sounds in order to understand what the speaker is saying. e. Metathesis. Rules of metathesis change the order of sounds. In many instances, sounds metathesize in order to make words easier to pronounce or easier to understand. In Leti, an Austronesian language, consonants and vowels switch places when a word that ends in a consonant is combined with a word that starts with two consonants. The last two sounds in the first word trade places to avoid haVing three consonants in a row. (16) CV metathesis (Leti): When three consecutive consonants occur, the first consonant trades places with the preceding vowel. By this rule, /danat + kviali/ 'millipede' undergoes metathesis to become [dantakviali}, and /ukar + ppalu/ 'index finger' becomes [ukrappalu]. On the other hand, /ukar + lavan/ 'thumb' does not undergo metathesis and so is pronounced as [ukorlavan] because there are not three consecutive consonants. Metathesis is also possible in sign languages. In these cases, the order of two signs or parts. of signs is switched. For example, in ASL, several signs are articulated with a path of movement that moves from a location just beneath the ear to a location by the side of the mouth. However, many speakers of ASL will produce these words with these two places of articulation reversed. Consider the signs for DEAF in (17). These two pronunciations include the same handshape, orientation, and type of movement; however, the order of the two locations is reversed in the second pronunciation. (17) a. DEAF (indexical form) 2006, William Vicars, www.Lifeprint.com. U sed with perl11ission. 2006, William Vicars, v,'WW.Lifeprint.com. Used with permission. Metathesis may similarly occur between places of articulation in signs with movement along a different path, for example, between forehead and chin or from one side of the chin to the other. Other signs in ASL that allow metathesis of place of articulation include RESTAURANT, FLOWER, PARENTS, and TWINS. f. Strengthenillg. (also called fortition) make sounds stronger. The rule of Epglish aspiration, .as stated below,provides an e"ample: (18) Aspiration (English): Voiceless stops become aspirated when they occur at the beginning of a stressed syllable. File 3.2 Rules 115 The pronunciation of tap Iteepl as [tha:p] and cat Ikeetl as [kha:t] illustrate the application of the English aspiration rule. Aspirated stops are considered to be stronger sounds than unaspirated stops because the duration of voicelessness is much longer in aspirated stops (since it extends through the period of aspiration). g. Weakening. Rules of weakening (also called lenition) cause sounds to become weaker. The "flapping" rule of English is an example of weakening. [f] is considered to be a weaker sound than [t] or Cd] because it is shorter and it obstructs air less. (19) Flapping (English): An alveolar oral stop is realized as [f] when it occurs after a stressed vowel and before an unstressed vowel. The pronunciation of writer 11QIql as [lQIq] and rider IlQld:p as [lQIq:] are examples of the application of this rule. Note that voicing assimilation is involved in1:l1e change of It I to [f]: the It I takes on the "voicedness" of the vowels surrounding it. 3.2.4 Multiple Rule Application To this point we have seen examples where one phonological rule applies. In reality there is often more than one change that occurs between a given phonemic form and a phonetic output. To illustrate this let's look at how plural nouns are formed in English. When you learned to write in English, you learned that the way to make most nouns plural is to add an <s>, which is usually pronounced [z]. There (ire actually three different phonetic forms of the English plural marker: [s], [z], and [az], seen in the words cats [keets], dogs [d:>gz], and foxes [faksaz]. We need only one phonemic form for the plural marker if we use two rules to derive the phonetic forms. Let's assume that the single phonemic form of the plural marker is I-z/, rather than I-sl or l-az/. 3 Why do words like fox, ditch, bush, orange, rouge, and maze have [-az] as their plural suffix instead of just keeping the phonemic form I-z/? The answer lies in the last sound of each of these words---'they all end in sounds that have a high-pitched hissing sound quality. These sounds are called sibilant consonants. The English sibilants are rtf, f, s, d3, 3, z]. Notice that the plural marker is also a sibilant I-z/. Because of the high-pitched hissing sound, it is very difficult to hear two sibilants that are next to each other. Try saying [faksz], [dltJZ], (blId3Z], etc., and you will get the idea. This difficulty is remedied by inserting a schwa between the two sibilants. (20) Schwa insertion (English): Insert [a] between two sibilants. The schwa insertion rule makes these plurals pronounceable by separating the two sibilants: [faksaz], [dltJaz], (blId3az], etc. But why do we still need two different plurals, [-s] and [-z], for words that do not end in sibilants? Try to pronounce [ka:tz] or [d:>gs], in which the voicing qualities of the final two consonants differ. You will probably find that it is difficult to produce a consonant cluster if one consonant is voiced and the other is voiceless. It is for this reason that the plural marker changes its voicing specification to match the sound it follows. (21) Voicing assimilation (English): I-zl takes on the voicing specification of the preceding sound. The voicing assimilation rule takes a phonemic form like Ika:t-zl and turns it into the pronounceable form [ka:ts]. With these two rules acting together, we can derive the plural 3This assumption is in fact correct for English; the arguments for picking this form, however, are beyond the scope of this discussion. 116 for any English noun (except, of course, for "special" or irregular plurals like oxen, octopi, or cherubim). In (22), you can see how these derivations work: starting with the phonemic form, we apply each rule in turn (first schwa insertion, then voicing assimilation), and we end up with the correct final phonetic form.. (22) Sample derivations of English plurals phonemic form: schwa insertion: voicing assimilation: phonetic form: krets [krets) IdJg-zl [dJgz) Ifaks-zl Ibnd3- z1 faksaz Ikret-zl bnd3az [faksaz) 3.2.5 Obligatory and Optional Rules Notice that phonological rules may be obligatory or optional. Obligatory English rules include aspiration, vowel nasalization, vowel lengthening, and liqUid and glide devoicing. Such a rule always applies in the speech of all speakers oia language or dialect having the rule, regardless of style or rate of speaking. The effects of obligatory rules are often very subtle and difficult to notice, but they are an important part ofa native accent. For instance, it may be difficult t() tell that a vowel is niisalize.d in English, but not applying the rule of vowel nasalization would make someone sound like a non-native speaker of English. The existence of obligatory rules is what causes people to h.ave foreign accents. It is easier to learn the rules ofa new language than to "turn off" the obligatory rules of your native language. The very fact that we are often unaware of these rules causes us to apply them when they are not appropriate. When speakers of American English learn other languages, they often apply rules such as flapping and vowel reduction, even though the other m ay not have these rules. Optional phonological rules, on the other hand, mayor may not apply in any given utterance. Optional rules are responsible for variation in speech; for example, we can pronounce Ikam bil can be as either [keem bi) or [keen bi), depending on whether alveolar stop assimilation is applied or not. The use of optional rules depends in part on rate and style of speech. 3 .3 Phonotactic Constraints and Foreign Accents 3.3.1 Phonotactic Constraints In every language there are restrictions on the kinds of sounds and sound sequences possible in different positions in words (particularly at the beginning and end). These restrictions can be formulated in terms of rules stating which sound sequences are possible in a language and which are not. Restrictions on possible combinations of sounds are known as phonotactic constraints. Languages generally prefer syllables made up of a consonant (C) first, and a vowel (V) second, but some languages allow a syllal;>le to begin with more than one consonant. For instance, English allows up to three consonants to start a word, provided the first is lsi, the second Ipl, Itl, or Ikl, and the third 11/, Ill, Ij I, or Iwl (see below). There is awide variety of syllable types in English, as illustrated in (1) (1) V ve vee veee a at ask asked ev eve evee eveee no not ramp ramps eev eeve eevee eeveee flew flute flutes crafts eeev eeeve eeevee eeevece spree spleen strength strengths Other languages, however, do not have such a large number of syllable structures, as the lists in (2) illustrate. (HebreweVee syllables are allowed only at the endofa word, and only if the final consonant is [tJ.) (2) Hawaiian ev V evee Notice that this means that Indonesi;m has clusters only in the middle of words; that is, there are no clusters initially or finally. Hawaiian does not permit clusters in any position. Meanwhile, even though Hebrew permits both initial and final clusters, it does not allow a single vowel to be a syllable by itself. Every language has its own set of permitted segmental sequences. We can investigate examples of restrictions on consonant sequences in more detail by considering some in a language we know very well-English. To start with, any consonant of English may occur initially (at the beginning) in words except for two: [3J and [1)J. While some speakers do pronounce these sounds in borrowed words such as Jacques and Nguyen, 117 118 no native English word begins with them. A large number of two-consonant combinations also occur word-initially, with a stop or fricative being followed by a liquid or glide: [bl] [reI] bring three [gl] glean [fl] fly [mj] [hj] music humor [kw] [sw] quick sweet In addition, [s] can also be followed by voiceless and nasal stops (as in stay, small) and by [f] and [v] in a small number of borrowed words (sphere, svelte, etc.). [f] can be followed by a nasal stop or a liquid, but only [fl] is a cluster native to English (e.g., shrink). The others are present only in borrowings from Yiddish and German (Schlemiel 'clumsy person,' Schnook, 'fool,' Schwinn). 3.3.2 Phonotactic Constraints in Signed Languages There are similar kinds of constraints on what sorts of segment combinations are and are not allowed in various signed languages. As with the phonotactic constraints for syllable structures and for consonants and vowels in spoken languages described above, constraints on syllable structure and on what sorts of handshapes and movements can appear adjacent to one another in signed languages differ from language to language. The phonotactic constraints discussed in this section are specific to ASL. First we will consider restrictions on syllable structurej there will be two examples.! It was mentioned above that in Hebrew, a vowel alone cannot serve as a syllable: there is a minimum requirement that a syllable in Hebrew contain at least two segments. There is a similar minimum requirement for ASL syllables: a monosyllabic sign cannot consist of just one handshape, one location, and one orientation; at least one of these elements is required to change in order to form a grammatical syllable. The second example we will consider is when changes of handshape are allowed. Many signs include a change of handshape during movement of the hands from one location to anotherjother signs involve handshape changes that occur while the hands are held stationary at some particular place. In ASL, handshape changes may always occur during movement. The sign WHITE, shown in (3), proVides a good example of this. (3) ASL: WHITE 2006, William Vicars, www.Lifeprint.com. Adapted by permission. Likewise, if a sign comprises only one place of articulation without movement to another place of articulation, handshape change can occur while the hand is kept at that location. However, if a sign involves the hands being at some place and then moving, or 1It probably seems very peculiar to think about signs as having syllables! Nonetheless, signs can be broken down into prosodic units just like spoken words. If you are not a signer, though, it is very difficult to figure out what might comprise a signed syllable. (We also lack intuitions about syllable structure of spoken languages that we do not know.) Therefore, we will simply take it as an underlying assumption that signed languages have syllables, and go from there. File 3.3 Phonotactic Constraints and Accents 119 moving and then winding up at some place, then the change of handshape must take place during the movement. It is ungrammatical in ASL for handshape to change while the hands are held at some particular location if there is a movement component of that sign. To illustrate what this would be like, try to mimic the sign for 'white' shown in (3). Place your hand flat on your chest. Next, draw your hand away from your chest (to a distance of about 5 or 6 inches), and as you do so, bring your thumb and fingers together. (You won't be forming a fist; rather, imagine that you are holding a small object between your fingers and thumb,)This sign is perfectly acceptable in ASL. "Now you are going to produce an ungrammatical Sign. Again, place your hand flat on your chest. Next, move it away from your chest, again about 5 or 6 inches-this time keeping your hand flat-ap.d once you have finished moving your hand, bring your fingers and thumb in to touch each other. Of course, there is nothing at all articulatorily difficult about making such a motion, but it is not a permitted sign in ASL; this constraint is merely an arbitrary rule built into the phonologyof ASL. Not only do signed languages have syllable structure constraints, but also there are constraints on which segments can be adjacent to one another in the same way that certain sound combinations are not allowed in some languages (like word-initiallfkl in English). For example, in any given sign language, there may be certain handshapes which-though parts of the system of handshapes in that language-are not allowed by the grammar to appear adjacent to one another within.a word. An interesting phonotactic constraint that does not have any parallel in spoken languages but that seems fairly uniform among sigp.ed languages pertains to the fact that in signed languages there are two possible primary articulators, namely, .the right hand and the left hand. In all signed languages studied to date, a signer maybe right-hand dominant or left-hand dominant (which roughly corresponds to whether the. signer is right- or left-handed). The dominant hand is the one that the signer will useto perform all one-handed signs. (If a right-handed signer were to injure her right arm orhappep.ed to be carrying a large box under her right arm, she might temporarily switch and use her left hand to sign with, but no signer switches back and forth between hands as a matter ofcourse: this would be considered extremely aberrant to any native speaker oia sign language.) Interestingly, there are very specific restrictions on what the non-dominant hand may do in any given two-handed Sign. If both hands are moving, then the non-dominant hand must have the same handshape and orientation as the dominant hanq and must move in the same way: that is, in signs where both hap.ds are. moving, there is a symmetry <;Oi).straint. (You can think of this as the "don't pat your head and rub youLPelly" An example of an ASL sign that follows the symmetry constraint is CAN, meaning 'be able to, , illustrated in (4). Although the non-dominant hand does move in this sign, it mirrors exactly the shape, orientation, and movement of the dominant hand. (4) ASL: CAN 2006, William Vicars, Used with permission. The non-dominant hand may also participate in a sign by remaining stationary while the dominant hand moves. This is exemplified in the ASL sign CHOCOLATE in (5): the two hands have different hand shapes, but the non-dominant hand (the lower hand, which is held flat) is not moving. (5) ASL: CHOCOLATE 2006, William Vicars, www.Lifeprint.com. Used with permission. A grammatical sign in any sign language cannot have both hands moving unless they both have the same handshape and orientation and are performing the same kind of movement. Interestingly, in Signed Mandarin (which is not a sign language, but rather a signed code for Mandarin Chinese; see File 1.5) there are certain signs that have been introduced by hearing (non-native signer) instructors at deaf schools that do not follow this rule. The fact that the Signed Mandarin words do not follow the universal rule for signed languages is yet more evidence that signed codes and signed languages differ! For example, in the signed Mandarin sign for 'ink,' both hands are moving, so the sign should follow the symmetry constraint. However, in this sign the dominant hand is facing toward the signer, and the non-dominant hand away from the signer, so they have different orientations; the dominant hand moves in a path away from the body while the non-dominant hand moves from side to side, so they have different movement; and the dominant hand has one finger extended while the non-dominant hand has three fingers extended, so they have different handshapes. Thus this sign is ungrammatical for three reasons. There has been an attempt (again by non-native signers) to introduce such signs from Signed Mandarin into Taiwan Sign Language, but the reaction among the native signers is that these signs are not possible in their language. It is exactly as though somebody told you that Ikpflusl was a new word of English; you wouldn't accept it! 3.3.3 Foreign Accents If a language has severe restrictions on its phonotactics, the restrictions will generally apply to every word in the language, native or not. Therefore, languages seek to overcome problems of borrowing a foreign word that violates their phonotactics. For instance, in English, two stops cannot come at the beginning of words, nor can stop plus nasal combinations. So, in order to pronounce the foreign words Ptolemy and gnostic more easily, English speakers simply drop the first consonant and pronounce the words [taldmi] and [nastIk], respectively. Or, speakers may insert a vowel between the two consonants, as in the pronunciation of the words Gdansk and knish as [gdd<ensk] and [kdmJ]. As these examples from English illustrate, there are different ways of handling phonotactic problems. Japanese and Finnish proVide us with additional examples. Japanese and Finnish generally avoid syllables containing sequences of consonants. When borrowing a foreign word that violates language's the syllable structure, the two languages must force it somehow to fit. There are two ways that borrowed words with consonant clusters are "repaired." One is to drop or delete one of the consonants; the other is to insert a vowel to separate the consonants. Finnish opts for deletion. In loan words, Finnish drops the first of a series of consonants that do not conform to its phonotactics. Thus, Germanic Strand (CCCVNC) ends up as ranta 'beach' (CVNCV) in Finnish, and glass becomes lasi. Note also the addition of a final vowel to avoid a consonant in syllable-final position. The other way to break up consonant clusters is used in Japanese. Japanese inserts vowels into the cluster, so that, for example, a CCC sequence will end up as CVCVCv. The insertions are rule-governed, meaning that the insertion always works the same way; the vowel lui is inserted, except after It I and Idl, when the vowel 101 is inserted. Thus, we can predict the shape of new words in Japanese (e.g., recent English loan words in the language). F ile 3.3 Phonotactic Constraints and ForeIgn Accents 121 So, for example, when the English term birth control was borrowed into Japanese, it became [ba:su kontoro:ru]. Note that the nasals [m] and [n] are allowed syllable-finally in Japanese. Iba181 IkantlOl! Iba:sul Ikontoro:rul The lui in [ba:su] and the last lui in [kontoro:ru] are inserted to keep the word-final syllables from ending in a consonant. The second [0] in [kontoro:ru] is inserted to prevent [t] and [r] from forming a cluster. Notice also the substitutions made by Japanese for English sounds: IVrl -t IV:/, 181 - t lsi, /1/-t Ir/. Another source of foreign accents is the fact that not all sound systems are the same, as we discovered in conducting phonemic analyses of different languages. Some languages have fewer or more phonemes or allophones than English does, and we can detect this when we hear non-native speakers of English pronounce English. For instance, French speakers often pronounce English this [OIS] as [ZIS] and thin [8m] as [sm]. The reason for this mispronunciation is that the phonemic inventory of French does not contain 101 or 18/, so French speakers substitute the nearest equivalent sounds, the fricatives Izi and lsi, available in their phonemic inventory. This is known as sound substitution, a process whereby sounds that already exist in a language are used to replace sounds that do not exist in the language when borroWing or trying to pronounce a foreign word. Another familiar example involves the pronunciation of German by some speakers of English. German has a voiceless velar fricative phoneme, represented by the symbol Ix/. English, of course, lacks this sound, though we do have a voiceless velar stop Ik/. Most speakers of English substitute Ikl for Ixl in a German word like Bach Ibax/, prodUcing [bak]. Another example of the same substitution is the way Americans will tend to pronounce the German word Lebkuchen Ilebkuxcml 'Christmas cookie' as [lelbkukan]. Some English speakers, striving for a more "Germanlike" pronunciation, will pronounce it instead as [lelbkuhan]. Why do you suppose an English speaker might substitute Ihl for Ix/? We can conclude by observing that substitutions by non-native speakers and strategies for handling phonotactic constraints both result in foreign accents, as well as changes in words that have been borrowed into another language. A Spanish speaker does not pronounce student as [estudent] because he or she doesn't know any better but because the consonant clusters Istl, Iskl, and Ispl never occur at the beginning of a word in Spanish without being preceded by a vowel-for example, in the words estudiante 'student,' escuela 'school,' and espalda 'shoulder.' The Spanish speaker who says [estudmt] is simply applying the phonological rules of Spanish when speaking English words. 3 .4 Implicational Laws 3.4.1 Recurring Phonological Patterns In studying phonetics, you saw that human languages use a wide variety of sounds. In spite of this variety, some sounds are more common than others. Thus, while it is true that almost all human languages use the stop consonants [p] and [tJ and the vowel [a], relatively few languages use pharyngeal fricatives ([nJ and [l], the "throaty" sounds used in Arabic), voiceless vowels (like in whispered speech), and clicks.(tsk, tsk! and horse calling sounds are American examples). So [p], [t], and raj are mOTe common in languages while pharyngeal fricatives, voiceless vowels, and clicks are less common speech sounds. The purpose of this file is to explain why some sounds are more common than others. Before attempting an explanation, however, we will make four observations concerning more common and less common speech sounds. 3.4.2 Sound Inventories The first observation has to do with the inventories of sounds in languages. The observation is basically this: if a language uses a less common sound, one of its more common counterparts will also be used. Two parts of this statement need clarification. First, when we say that a language uses a particular sound, we mean that the sound is in the inventory of phonemes in the language. In other words, that sound is distinctive relative to other sounds in the language. The second part of the statement that needs clarification is the phrase "one of its more common counterparts." This phrase refers to the fact that for each less common sound in the inventory there tends to be a more common sound in the inventory which is just like the less common sound except far one or two phonetic features. For instance, the more Common counterpart of a voiceless vowel is a voiced vowel of the same tongue height, tongue advancement, and lip rounding. likeWise, the more common counterpart of a voiceless pharyngeal fricative is a voiceless velar fricative. The table in presents some (relatively) less common sounds and their (relatively) more common co'unterparts. One to notice about this chart is that [s] appears both as a more common sound (as opposed to and as a less common sound (as opposed to [t]). This illustrates the fact that in using the terms "more common" and "less common" to designate the sounds in an implicational relationship, we are not referring to an absolute standard. Rather, "more common" and "less common" are used in a relative way. In other words, [s] is less common in relation to [t], but more common in relation to [xl. We have that if a language uses a less common sound, one of its more common counterparts will also be induded in that language's inventory of distinct sounds. In terms of the chart presented in (I), this means that any language that uses [a] will also use [a], any language that uses will also use [a], any language that uses [x] will also use [k], and so on. This type of observation is called an implicational law because the presence of the less 122 File 3.4 Laws 123 (1) Less common More common [a] [a] [a] [x] [s] [d] [k] or [s] [5] voiced stops fricatives in place X [t] [t] [d] or [z] voiceless stops stops in place X common sound implies that the more common sound will also be used in the language. Of course, the implication cannot be reversed. In other words, the fact that English uses the sound [k] does not imply that we also use [x]. Implicational laws can be stated for natural classes of sounds rather than just for individual pairs of sounds. For instance, the class of voiceless consonants is relatively more common than the class of voiced consonants. In other words, if a language makes use of voiced stops, it will also make use of voiceless ones. The reverse is not true; there are some languages that have only voiceless stops. Thus, the presence of voiced stops implies the presence of their voiceless counterparts, while the presence of voiceless stops does not imply the presence of voiced ones. Another implicationallaw that can be stated in terms of a natural class of sounds is that the presence of fricatives in a language implies the presence of stops with the same place of articulation as the fricatives in that language. Thus, if a language uses an [s], then it also uses a [t]. 3.4.3 Frequency and Distribution The second observation concerning implicationallaws is that they are not simply generalizations concerning inventories of sounds; they are also related to the t o which sounds will be used in a particular language and to the range of distribntion of the sounds in the words of the language. Thus, even if a language makes useofa pharyngeal fricative, this less common sound will be used in fewer words than wilLthe more common velar fricative. In other words, the pharyngeal fricative will have limited usage compared with the velar fricative. More common sounds have a wider distribution within a languageLe., they are used in more phonetic environments than less common sounds. So, for instance, Cantonese Chinese has both stops and fricatives in its inventory of sounds, but fricatives may occur in only one position inthe syllable, as the first sound. Stops have wider distribution: they occur both syllable-initially and syllable-finally in Cantonese. An English example of the limited usage and limited distribution of less common sounds has to do with the sound [5]. The sound [5] can be classified as less common because it is relatively rare in the languages of the world, and anywhere [5] occurs in English, [z] can also occur. If you try to think of words that contain [5], you will probably find that your list is limited to grammatical" words like this, that, those, them, and they, and a few other words like either and lathe. Furthermore, [5] occurs as the last sound in English words less often than [z] does. Compared with the number of words that contain [z], it is obvious that [5] has limited use in English. (I 124 3.4.4 Acquisition of Sounds A third type of observation related to implicational laws has to do with the order of the acquisition of sounds: children learning a language acquire the use of more common sounds before they acquire the use of less common ones. As a result, children who have not yet mastered the complete sound inventory of their native language will substitute more common sounds when trying to say less common sounds. When a little girl says [dIS WAn] for this one, she is replacing the relatively less common [5] with [d], a much more common sound. This is an indication that the child has not yet fully acquired the use of [5], although [d] is readily available for use. When the language development of a child is followed from babbling through maturity,.a characteristic order 9f acquisition appears. This order in the acquisition of sounds is relatively constant for children around the world, no matter what language they are learning. Once again, the implicational laws capture a generalization about language; namely, that the acquisiti9n of arelatively less common sound implies that its more common counterpart has already been acquired. 3.4.5 Sound Change The fourth and last type of observation related to implicational laws involves language change: less common sounds tend to be less stable than more common ones. Thus, in the coutse oflanguage change, if any sound is going tobe lost, it is more likely to bea less common one rather than its more common counterpart. An illustration of this can be drawn from the history of English. In the Old English pronunciation of the word knight there was a voiceless velar fricative [x] between the vowel and the [t]. As you can see, the letters <gh> indicate where this consonant used to be. During the development of English, this velar fricative waslost (so knight now rhymes with quite). In fact, all instances of the velar fricative sound (as in height, sight, fight, might, and so on) were lost. English speakers just stopped using velar fricatives altogether, so noW we find it hard to learn how to say them when we are trying to learn a language like German that uses them. This observation fits in with the implicationallaw that says that fricatives are less common than stops. Therefore, the fricative [x] is less stable and more likely to be lost than the corresponding stop consonant [k]. For more on sound change, see File 12.3. 3.4.6 Implic;}tional Laws At this point we can summarize what we have observed about common and uncommon speech sounds: the presence of a less common sound in a language implies that its more common counterpart will also be present; less common sounds have limited usage and distribution in the languages that do make use of them, as compared with common sounds; the use of common sounds is acquired before the use of less common ones; and less common sounds tend to be less stable than common ones and are thus more likely to be lost or changed over time. We might be tempted to say that the implicationallaws are themselves the explanations ofthe observations. Thus, we might say that [x] is more likely to be lost in language change than[k] is because [k] is more common than [x]. Or we might want to say that [k] is acquired by children before [x] because [k] is more common than [x]. This type of explanation is circular, however. The circularity stems from the fact that we distinguished between common and less common sounds by making the observations. The alternative to this circular form of explanation is to explain the above observations (and thus the implicationallaws) in terms of the communicative nature of language. It is important to realize that when people use language, their goal (generally speaking) is to communicate-that is, to successfully transmit a messagetrom a speaker toa hearer (refer File 3.4 Laws 125 t o diagram (1) in File 1.2). Focusing on the function of language leads usto ask what sounds are most useful for transmitting a message from speaker to hearer. First of all, notice that if a sound is difficult to produce, speakers will be somewhat inconsistent in pronouncing it,and this inconsistency may result in confusion on the part of the hearer. To avoid being misunderstood, speakers may avoid words with difficult sounds (resulting in limited usage), and if enough speakers avoid a difficult sound, it may disappear from the language entirely (language change). Of course, sounds that are difficult to produce (such as fricatives, whose production involves delicate control of muscles) are not likely to be mastered by children before easier sounds are. As you can see, there are at least some instances where the observation that sound X is more common than sound Y is directly tied to the fact that sound X is easier to produce than sound Y. Thus, [k] is more common than [x] because stops are easier to produce than fricatives. Alveolar fricatives are more common than pharyngeal fricatives because the tip of the tongue is more agile than the back of the tongue; hence alveolar consonants are easier to produce than pharyngeal ones. Thus, ease of production is an explanation of at least some of the implicationallaws. Another way to answer the question of what sounds are most useful for transmitting a message from speaker to hearer focuses on the hearer's point of view. It is reasonable to suppose that if a sound blends into the surrounding sounds too much, its distinctive qualities may become difficult to hear. So, for example, if Morse code were made up of long dashes and not-so-Iong dashes, or dots and somewhat shorter dots, rather than dots and dashes, it would be difficult to use. In the same way, the consonants and vowels which make up syllables are most usable when they are quite different from each other. So, the kind of syllable which is most useful in transmitting messages in language is composed of maximally distinct consonants and vowels. By this we mean that the consonants have very few qualities in common with the vowels, and the vowels are likewise very different from the consonants. The value of maXimally distinct carriers of information is obvious when we think about Morse code. If you can't tell the difference between dots and dashes, then little communication can take place. In the same way, if you can't tell the difference between consonants and vowels, then communication using language is likely to be very inefficient. Perhaps a couple of examples of the ways that consonants can be more vowel-like, or vowels can be more consonant-like, are in order. One implicationallaw that we noticed is that the use of voiced consonants in a language implies the use of voiceless ones (thus voiceless consonants are more common than voiced ones). The natural explanation for this implicationallaw is that voiceless consonants have fewer qualities in common with vowels than do voiced consonants; thus, in syllables containing consonants and vowels, voiceless consonants are perceptually more salient (or noticeable) than voiced ones. A way that vowels can be less consonant-like is to be pronounced with the mouth Wide open, as in the vowel [a]. Because consonants are made by obstructing the vocal tractin some way, a vowel that is pronounced with the mouth wide open will be more distinct from surrounding consonants than will be a vowel like [i] or [u] which is pronounced with the mouth somewhat closed. It just so happens that there is an implicational law corresponding to this distinction between [i], [u], and [a]. The presence of a closed vowel ([i], [ul) implies the presence of an open vowel ([a]). Thus, syllables with maximally distinct consonants and vowels are easier to perceive than syllables with consonants and vowels that resemble each other, and therefore some implicational laws exist for the sake of the listener, to make language easier to perceive. 3.4.7 Conclusion In this file we have seen that although there is great variety in the sounds that can be employed in language, there are universal tendencies: to restrict the inventory of sounds to 126 certain more common sounds, to restrict the degree of utilization and distribution of less common sounds in languages that do use them, to acquire more common sounds earlier than less common ones, and for less common sounds to be unstable in the face of language change. We have also shown that these observations concerning more common and less common sounds are related to the ease of production and ease of perception of those sounds. In addition, the implicationallaws can at least sometimes be explained by assuming that people are using language in order to communicate, and that this produces a need for efficiency which leads to the use of easily produced and perceived sounds. 3 .5 How to Solve Phonology Problems 3.5.1 Goals of Phonological Analysis Because phonemes are important units of linguistic structure, linguists must have a general method for identifying them in all languages. But the task of determining what the phonemes of a language are and what allophones are assigned to them is not always straightforward. For one thing, the set of phonemes differs from language to language, and so a different analysis is required for each language. Moreover, phonemes are psychological units of linguistic structure and are not physically present in a stream of speech. As a result, it is not possible to identify the phonemes of a language simply by taking physical measurements on a sample of Nor is it always easy to identify investigating a native speaker's intuitions, since the minute phonetic details on which.decisions about phonemes are made are often precisely those which speakers are not accustomed to noticing. To get around these problems, linguists have developed an objective by which the phonemes of a language can be discovered through examination of a set of words written in phonetic transcription. This procedure is based on two main observations about patterns of sounds. First, phonemes make distinctions in meaning. If two sounds are members of separate phonemes, minimal pairs can almost always be found. For example, the minimal pair led and red is evidence that [I] and [1] are members of p honemes in English. But if two sounds are allophones of the same phoneme, minimal pairs differing only in those sounds will not exist. For example, [bA?l,l] and [bAthl,l] are both possible pronunciations of the English word button (though [bAthl,l] may sound a little stilted). This is because the sounds [?] and [t h] are both allophones of the phoneme It/. Thus, the meaning doesn't change. Second, the allophones of a phoneme are not a random collection of sounds but are a set of sounds that have the same psychological function. Accordingly, allophones of the same phoneme are systematically related to one another: they often share many phonetic properties, and it is possible to predict which allophone will appear in a word on the basis of phonological rules. By analyZing the patterns of sounds that are physically present, it is possible to draw conclusions about the psychological organization of a language, which is not directly observable. 3.5.2 How to Do a Phonemic Analysis Although a phonemic analysis can be performed successfully on any language, it is easiest to begin with a problem based on English. Look over the data in (1), which are given in a fairly detailed phonetic transcription. Recall that an open circle under a segment indicates that it is voiceless. 127 128 (1) 'pray' 'gray' 'crab' 'pa.r' 'broker' [phi eI) [g1eI) [kh{eeb) [phOl) [blOukt ) 'fresh' 'regain' 'shriek' 'tar' [1igem) [thOl) Beginning with the sounds [1) and we attempt to answer the following question: are these sounds allophones of separate phonemes, or allophones of the same phoneme? (Of course, native speakers of English may intuitively know that they are allophones of the same phoneme. However, the procedure for doing a phonemic analysis should produce the same answer without appealing to the intuitions of speakers.) In order to answer this question, it is necessary to examine scientifically the distribution of sounds within these data. That is, for each sound in question we need to determine the set of phonetic environments in which it can occur. But just what do we mean by environment? For the time being, we can define the environment of a sound as the sounds that immediately precede and follow it within a word. For example, in the word [g1eI), [1) is in the environment [g_eI); that is, [1) is preceded by [g) and followed by [eI). The best way to begin a phonemic analysis is to determine whether the sounds in question are contrastive. To do this, look first for minimal pairs. Suppose for a moment we were interested in the sounds [ph) and [th) in the data in (1). These sounds do appear in a minimal pair: [phOl) and [thOl) have different meanings and differ phonetically by only a single sound in the same position. This tells us that [ph) and [t h) are in overla.pping distribution and, more specifically, that they are in contrastive distribution, because the difference between them causes a difference in meaning. Therefore, they are allophones of different phonemes. Returning to the problem at hand, namely, the status of [1) versus [i), we see that there are no minimal pairs in the data that differ only by these two sounds. Since [1] and [i) are not in overlapping distribution in our data, l we can assume that they are in complementary distribution. However, we must prove that this is so by making a generalization about where [1] (but not [iD may appear, and vice versa. In order to do sO,we need to compare the phonetic environments of each of these sounds. The easiest way to do this is to make a list for each sound, as follows. (Note that 1/#" indicates a word boundary.) (2) [1) [g_eI) [<L#) [b_ou) [#-iJ [i) [ph _eI) [kh _ee) [eEl [J_i) Once you have collected the list of phonetic environments for each sound, you can proceed as follows: 1. Look at the environments to find natural classes. [1) is preceded by [ph), [kh), [f), and [J), all of which are voiceless consonants. This generalization permits us to simplify the description of the environment for [i); instead of listing each sound separately, it is now possible to say: l you can always assume that the data you are given are representative of the language pattern you are asked to analyze for the purposes of solving phonology problems in this book. Sometimes we have selected a particular subset of the data from a language to illustrate a particular analytical point; this should not be taken as a sign that every word in the language will follow exactly the same pattern. However, the patterns we present are representative of basic phonological distributions. File 3.5 How to Solve Problems 129 (3) [.!J appears after voiceless consonants. Now look at the environments in which [1] appears. Are there any natural classes? Yes and no. Certainly [b] and [9] are voiced consonants, and [0] is also voiced, but the set that includes [bJ, [9], [oJ, the beginnings of words, and the ends of words does not form a natural class. Thus, the critical observation to make here is that there is no single natural class of environments in which [1] can be found. We have looked at the sounds preceding [1] and b ut what about the sounds that follow them? As you can see, only [1] may occur word-finally, but either [1] or c an occur before a vowel. Because the environment that follows either [1] or [-1] can be the same (for Thus, the environexample, [ell) this alone can't tell us about when you get [1] versuS ments that condition the appearance of [1] or [-1], Le., the conditioning environments of these particular allophones, are their immediately preceding sounds. 2. Look for complementary gaps in the environments. So far, we have shown that appears after voiceless consonants, while [1] appears in an apparently random set of environments. Yet, it is possible to make one more critical observation. [1] does not appear in the environments in which appears, namely, after voiceless consonants. Moreover, [-1] does not appear where [1] doesi there is no [-1] after voiced consonants or at the beginnings or h ave systematic and complementary ends of words. Since the environments of [1] and gaps, we say that [1] and are in complementary distribution. Therefore, they are allophones of the same phoneme. 3. State a generalization about the distribution ofeach ofthese sounds. In other words, write a rule that will make predictions about where each of the sounds can occur. Actually, we've done the hard part of this already by observing that [-1] occurs following voiceless consonants. How should we state the distribution of [1]? We could try formulating our rule as follows: (4) appears follOWing voiceless consonantsi [1] appears follOWing voiced consonants or vowels, or at the beginning or end of a word, However, that's not a very succinct formulation of the rule. To simplify it, recall that wherever occurs, [1] can't, because their possible environments form complementary sets. Therefore, we can revise our rule this way: (5) appears follOWing voiceless consonantsi [1] appears elsewhere. 4. Determine the identity of the phoneme and its allophones. This next step in writing the rule involves deciding what the phoneme to which these sounds belong should be. In order to do so, we need to decide which of the allophones is the basic allophone and which is the restricted allophone. We have determined that the conditioning environment for [-1] consists of a single natural class of sounds. is restricted to occurring onlyth.ere, whereas [1] may appear anywhere else. Therefore, we can identify [-1] asthe restricted allophone and [1] as the basic one. It makes sense to name the phoneme after the basic allophone, since it is the one that can show up in a wider variety of contexts. Furth.ermore, the basic allophone is assumed to be the closest approXimation of the mental "sound" that speakers store in memory. In choosing a name for the phoneme, we have made the leap from observable phonetic reality to unobservable psychological reality. {It is not always possible to choose one allophone as basic, however. In that case the phonology exercise's instructions will 130 not tell you to do so, and any of the allophones would serve equally well as the name of the phoneme.) We can improve on our rule once more by writing it to show the process of going from the phoneme to each of the allophones, as in (6). This notation was introduced in Section 3.2.1. The arrows in the rule in (6) mean 'is pronounced as.' We use slashes around symbols that represent phonemes, and a single slash indicates the beginning of the environment specification: (6) III --t I after voiceless consonants; III --t [1] I elsewhere. Now that we have formulated the necessary phonological rule, we can see which phonological process it involves (d. File 3.2). In this rule a voiced phoneme changes into a voiceless sound when it follows another voiceless sound. In other words, /JI becomes more like a .preceding sound with respect to the feature of voicelessness. Therefore, we can conclude that the process of assimilation is involved in this phonological rule. 3.5.3 Some Potential Trouble Spots The procedure outlined in the previous section will work for any language for which reliable phonetic transcriptions exist. However, beginners are often confused by certain questions. For instance, if you discover that no minimal pairs exist for two sounds, is it possible to automatically conclude that they are allophones of the same phoneme? No. It is still necessary to show that the sounds arein complementary distribution, since allophones are predictable variant pronunciations of the same phoneme. Consider what happens if you make a decision too soon. Using the data presented in (1) at the beginning of the previous section, suppose you wanted to know whether [g] and [Il are allophones of the same phoneme. Since there are no minimal pairs differentiated by these sounds in the data set, it might seem reasonable to conclude that they are. (Of course, a speaker of English should have no trouble thinking of a minimal pair involving these two sounds, for example, gag and gash. The exercises, however, are designed to be selfcontained; that is, in all of the problems in the book, you will be given enough data within the problem set to solve the problem. This means that you should not rely on outside knOWledge you may have of the language you are analyZing to answer the question.) But a careful examination of the data reveals that this is the wrong conclusion. Listing the data and the relevant environments, you find what is shown in (7). (7) [g] appears in gray [gleI], regain [ligem] generalization: [9] appears between vowels or at the beginning of a word; [J] appears in fresh [frEf], shriek [frik] generalization: [f] appears at the beginning or end of a word. As these data illustrate, [g] and [J] are not in complementary distribution because their distributions overlap: either may occur at the beginning of a word. Furthermore, either may be followed by the phoneme Ill. As a result, no phonological rule can be responsible for their distribution. In general, when no generalization can be made about where a group of sounds can occur, it is possible to conclude that they are members of separate phonemes. A conclusion based on such a demonstration is just as valid as showing that minimal pairs exist. This alternative way of shOWing that sounds are members of separate phonemes is useful because it's not always possible to find minimal pairs for all distinctive sounds. For example, there are no minimal pairs involving (1)] and [h] in English. But it is reasonable to File 3.5 How to Solve Problems 131 assume that they belong to separate phonemes because they share few phonetic properties, and no phonological rule determines where they can occur. The range of tests for identifying phonemes can be broadened somewhat by the use of near-minimal pairs. Recall that a minimal pair is a pair of words differing in meaning but phonetically identical except for one sound in the same position in each word. The definition of near-minimal pairs is the same, except that the words are almost identical except for the one sound. For example, heard a nd Bert form a near-minimal pair involving [h] and [b]. We are justified in saying that [h] and [b] are allophones of separate phonemes because no conceivable phonological rule would permit only [h] at the beginnings of words ending in [d], and only [b] at the beginnings of words ending in [t]. ([his conclusion is partly based on extensive study of how phonological rules work: experience does playa role in being able to do phonological analysis.) One final point about minimal pairs: notice that we have not defined them as pairs of words that rhyme. It is not necessary for two words to rhyme in order to form a minimal pair. Consider the English minimal pairs state [stert] and steak [sterk], for example, or boat [bout] and beat [bit]. Nor is rhyming sufficient to qualify a pair of words as a minimal pair: gray [gler] and pray [phter] from the list of data above rhyme, but differ in two sounds. And to take another example, glitter and litter rhyme but do not form a minimal pair because they do not contain the same number of sounds. Another question that often troubles beginners is this: when describing the environment in which a sound appears, how do you know where to look? In the problem we solved in the previous section, we considered only the sounds that preceded [1] and This is certainly not the only possibility. In fact, identifying conditioning environments is the most challenging part of doing a phonemic analysis. Recall that in many cases, the relevant conditioning environment consists of the sounds immediately surrounding the sound in question. However, it is sometimes necessary to look beyond the sound's immediate environment. As we saw for Finnish vowels in Section 3.2.3, if you are examining the distribution of a vowel allophone, it is quite common that the conditioning environment involves a vowel in an adjacent syllable, even though consonants may intervene. It may also be necessary to consider preceding or following sounds even when they belong to another word that is adjacent in the stream of speech. However, it is best to start by examining the immediate environment of an allophone when you are trying to determine what its conditioning environment is. Since there are many logically possible environments to consider, the task is made easier by eliminating all of those except the most plausible. This can be accomplished by using strategies like the following: a. Formulate hypotheses about the allophones. Investigation of the world's languages has revealed that some sounds are more common than others (see File 3.4 for relevant discussion). For example: Voiced nasals and liquids are more common than voiceless ones. Oral vowels are more common than nasal vowels. Short consonants are more common than long consonants. "Plain" consonants are more common than those with secondary articulations like velarization, palatalization, and labialization. On the basis of these generalizations, it is possible to speculate that if a less common sound appears in a language, it is probably a restricted allophone. But these tendencies should be used only as a guide for forming hypotheses, not as a basis for jumping to conclusions, since some languages exhibit exceptions. For example, French has both nasal and oral vowel phonemes. 132 b. Keep in mind that allophonic variation results from the application of phonological rules. Also remember that rules usually involve.some phonological process, such as assimilation or deletion. Once you have a hunch about which allophone is the restricted one, check the environment in which it appears for evidence that a phonological process has applied. This may involve looking in more than one place until you have discovered a reasonable candidate. In the problem in the previous section, we were guided by the knowledge that voicing differences in consonants are often caused by voicing assimilation, and that voicing assimilation frequently occurs in consonant clusters. Since /1/ is the second member of all of the clusters given, we conduded that the consonant preceding it constituted the conditioning environment. Even if it is not obvious that a phonological process has been at work, you should be able to write a phonological rule and, thus, state a generalization about where the allophones of the phoneme occur. 3.5.4 Flowchart for Discovering the Distribution of Sounds The flowchart in (8) should help you to identify the type of distribution two (or more) sounds in a language have. The rectangular boxes ask you to do something or give you some information that your working through the flowchart has revealed. The diamond-shaped boxes pose a question. Try reading through the flowchart before you attempt to analyze the languages in the next file (File 3.6, "Practice"); it may help you to understand the relationship between the different types of distributions of sounds in a language. F ile 3.5 How to Solve Problems 133 (8) A flowchart for identifying the distribution of sounds List the phonetic environments in which the sounds in question appear. Do the sounds occur in the same environment? Yes The sounds are in overlapping distribution. No The sounds are in complementary distribution. prediction is possible. Do the words have the same meaning? Yes Prediction is possible. No The words form a (near) minimal pair. The sounds are in contrastive distribution. 3.6 Practice File 3.1-The Value of Sounds: Phonemes and Allophones Exercises 1. Ukrainian Look at the following Ukrainian words containing the sounds [s), [Si], [f], and [Pl. The sounds [Si] and [fj] are palatalized variants of [s] and [f]; see the discussion in Section 2.4.6. The words have been arranged to help you identify minimal pairs. [s] a. b. c. d. e. f. g. h. [sj] [Irs] [mrska] [sapka] [sIla] [sum] [sudr] [sosna] [posadu] [lIsi ] 'fox' 'bowl' 'little hoe' 'strength' 'sadness' 'trials' 'pine' 'job' (ace.) [sjudr] [sjomrj] [posjadu] [f] 'sheen' 'hither' 'seventh' 'I will occupy' [p] [lIn [mrfka] [fapka] [fIla] [fum] 'lest' 'little mouse' 'hat' 'she sewed' 'rustling' [fostlj] 'sixth' [mrfii] 'mice' [f jisti ] 'six' [kopi] 'baskets' i. What minimal pairs can you identify in these words? ii. Is there a minimal triplet (like a minimal pair, but involving three sounds and three words)? What is it? iii. Which three of these four sounds are in contrastive distribution? iv. One of these sounds occurs only before a particular vowel. What is this sound, and what is the vowel? Which words indicate this? 2. Refer to the For Better or For Worse cartoon at the beginning of this chapter. This cartoon hinges on the use of a minimal pair, copulation versus population. What two consonants differentiate these words? Using your knOWledge of consonant articulation (see File 2.3), what are the phonetic descriptions of these two sounds? How do they differ phonetically? Discussion Questions 3. We have said that both contrastive distribution and free variation involve a context where it is impossible to predict which of two or more sounds belongs. However, these two are not the same thing. Consider the context [phla-l. Of the sounds [p, p', b, t, e, d), only one doesn't form a real word of English when inserted in this context-which one? Of the rest of these sounds, which ones are in contrastive distribution? Which ones are in free variation? How do you know? 134 File 3.6 Practice 135 4. Fill in the following table using the three terms "contrastive distribution," "complementary distribution," and "free variation" as defined in this chapter, with respect to two sounds in a given context. (For example, the upper left-hand cell of the table should contain the name of the type of distribution that occurs when two sounds are contrastive and predictable in a certain context.) Which cell in the table is blank? Why is it blank? Predictable Nonpredictable Contrastive Non contrastive 5. Refer to the For Better or For Worse cartoon at the beginning of this chapter. Elizabeth has clearly misunderstood the word population as copulation. Given what you know about the sounds [p] and [k] in English and the notion of a minimal pair, why does this challenge the notion of what it means for a pair of sounds to be "contrastive"? Do you think this is a serious challenge that undermines the phonological understanding of contrast? Why or why not? Activities 6. Obtain a dictionary or textbook for some signed language from your library, or go to an online sign language dictionary. (You may use a search engine to find one, or you may go to www.Lifeprint.com. which is the Web site of the man whom most of the ASL pictures in this book came from.) Look through the lists of words and try to find minimal pairs. The two words that you select should be the same in three of the following parameters and should differ in only one: place of articulation, movement, handshape, and orientation. For each minimal pair, specify which parameter the two signs differ in, and describe the difference. File 3.2-PhonologicaI Rules Exercises 7. List the members of the following natural classes of English sounds. a. alveolar obstruents b. voiced labial consonants c. velar oral stops d. interdental fricatives e. high tense vowels f. low vowels g. palatal sonorants h. voiced sibilants 8. Describe the following natural classes of English sounds. a. [I, 1] b. [f, 8, s, f, h] c. [w, i, ':Y J d. [i, u] e. [p, b] f. [n , I, 1] 136 9. Identify the phonological rule or rules from this file operating in each of the following derivations. a. little b. late bell c. park d. lance e. it's her car f. ten pages g. two cups /ht\! /leIt bel! /pmk/ /lams/ /Its h f km/ /tm peId3z/ /tu kApz/ [lull [leIp bel] [phmk] [lamts] [Its -1 khm] [tern pheId3az] [thu khApS] -7 -7 -7 -7 -7 -7 -7 10. Examine the following sets of data, and for each set write a rule to describe the derivation of the phonetic forms from the phonemic ones. (fo do so, determine what sound or natural class of sounds is being altered, what it is being changed to, and what the environment is. That is, your rules should be of the form X -7 Y/C_D.) Where possible, also explain what kind of process (of the seven types) is involved in the rule. a. In the speech of some New Yorkers, examples like the following are found. there court large stOres cared -7 marry Paris for all story caring [5e] -7 /5e1/ /kJ1t/ [btl [lod3] [sbZ] [ked] /IOld3/ -7 /St:>lZ/ -7 /k ld/ -7 /m<e1i/ /p<ens/ -7 -7 [m<e1i] [p<ens] /b1:>l! -7 [b1 :>1] /st:>li/ -7 /kElItJ/ -7 [st:>li] [kWI)] b. Examples like the following are very common in English. OSU did you capture gracious /oesju/ /dIdju/ /k<eptt/ / gleIsias/ -7 [oefu] -7 -7 [dId3U ] [k<eptJt] -7 [gleIfas] c. The following data are from German. (The symbol W/ represents a voiceless alveolar affricate.) German Gloss BUd blieb Weg fremd gelb Zug Vogel Baum schnell 'picture' 'remained' 'way' 'foreign' 'yellow' 'train' 'bird' 'tree' 'fast' /bIld/ /blib/ /veg/ /fremd/ /gdb/ /tug/ /fog\! /baum/ /Jnd/ -7 -7 -7 -7 -7 -7 -7 -7 -7 [bIlt] [blip] [vek] [fremt] [gelp] [t'uk] [fogH [baum] [Jnel] Discussion Questions 11. Consider the following paragraphs and answer the questions about natural classes. i. The English indefinite article is a [a] before most words: a car, apeanut, a tennis ball, etc., but it is an [<en] before words like apple, onion, icicle, evening, eagle, and honor. To what natural classes do the sounds at the beginning of each set of words belong? (That is, before what class of sounds do you use [a]? [<en]?) Ii. Some American English speakers (largely in the Midwest and the South) pronounce [I] in words like then, Kenny, pen, Bengals, gem, lengthen, Remington, and File 3.6 Practice 137 temperature (where other speakers have [cD. But, like others, they have [e] in words like pet, bell, peg, and tech. What natural class of sounds follows the vowel in words where these speakers have [I]? iii. Some midwestern American speakers in casual speech drop the unstressed vowel in the first syllable of words like police, believe, parade, Columbus, pollution, terrific, and collision, but do not drop it in words like detective, dependent, majestic, or pedantic. What natural class of sounds follows the unstressed vowel in the first syllable in the first group of words? iv. At some time during.a child's language development, he or she might pronounce certain words as follows: that [deet], these [diz], this [dIS], and three [bi], think [fIIJk], bath [beef]. What natural class of sounds is being affected? Do the sounds used as replacements form a natural class? 12. In File 3.2, two rules were used to derive the three phonetic forms of the English plural. Is the order of these rules important? Show what would happen if we applied the voicing assimilation rule before the schwa insertion rule. Give examples. 13. The traditional sign for TOMATO in ASL involves one hand with the index finger extended, moving from the lips down in front of the body, while the other hand is in a flat 0 handshape and remains still in front of the body. Some signers now produce it without the 0 hand shape, instead extending the index finger on the hand that stays still. What type of phonological process is this? Why do you think such a change might have happened? File 3.3-Phonotactic Constraints and Foreign Accents Exercises 14. According to the phonotactic constraints on English syllable structure given in Section 3.3.1, is [bljust] a possible word in English? Why or why not? Does this match with your own intuition? 15. List threedifferentways.an s peaker name Szczebrzeszynie [ftfeb3efl:]1e]fitinwith En:glish phOllotactics. Polish place Discussion Questions 16. If a consonant automatically be considered a legal phonotactic sequence in language? For do you think that [jl] should be considered phonotactically legal in English because it occurs in the words schlep and schlocky? or not? phonotacUcs of another lanthan just proo f the foreign to introduce 18. new i, Why do you think that the hearing instructors came up with signs that violate a universal principle of signed languages? If it is a universal principle, then why didn't the instructors create signs that followed it? 138 ii. Now that they have been created, do you think that these signs will catch on in TSL? Why, or why not? iii. Do you think that if they do catch on, they will be modified in any particular way, or do you think that they will keep the same form that they had in Signed Mandarin? (Be sure to refer to the information about foreign accents also discussed in File 3.3.) If you think they will change, what are some possible sorts of changes that could take place? iv. What do you think about the Mandarin-speaking instructors creating new signs? In general, should people who are not speakers of a particular language be allowed to introduce new words into that language? Is this case different, since the creators of the new signs have a native language with a different modality? File 3.4-Implicational Laws Exercises 19. Explain why it doesn't make sense to ask the question, "Is [s) a common sound in the world's languages?" 20. Given the explanations for implicationallaws given in Section 3.4.6, why do you think that clicks are relatively rare in the world's languages? Do you think it is related more to production or to perception? Why? 21. The explanations for implicationallaws given in Section 3.4.6 have also been used to explain other phenomena, especially in the domains of language variation and change. Look at the following pictures of the ASL word LEARN. One set shows the formal version of the sign; the other shows a more casual version. Speculate as to why the sign might have changed from the formal version to the informal version, given considerations of perception and production. a. ASL LEARN (more formal register) File3.6 Practice 139 Discussion Questions 22. Referring to the phonotactic constraints on syllable structure in File 3.3, do you think that there is an implicational hierarchy of syllable types? If so, what do you think it might look like? If not, why not? File 3.5-How to Solve Phonology Problems The exercises for this file are designed to give you practice in doing phonemic analysis at the beginning, intermediate, and more advanced levels .. The instructions to each exercise are somewhat different in each case, so read them carefully before proceeding. However, each exercise requires that you follow the step-by-step procedure for doing a phonemic analysis outlined in the text of File 3.5. The exercises are designed to introduce you to problems involving minimal pairs, complementary distribution, and free variation. A linguist doing a phonemic analysis of an unknown language would, of course, examine hundreds of words in order to be sure to have enough data to find the relevant minimal pairs, complementary distributions, etc. But to save you time, the data in the exercises below have been carefully selected to give you all the relevant information you will need in a very small set of words. Exercises 23. M okilese Mokilese is an Austronesian language of the Malayo-Polynesian ffmily, spoken in Micronesia. Examine the distribution of the voiced and voiceless vow pairs: [i, nand :l [u, I}] (voiceless vowels have a circle under the phonetic vowel symbol). For each pair, determine whether they are allophones of different phonemes or allophones of the same phoneme. Provide evidence for your answer. If they are allophones of one phoneme, state the contexts in which each sound occurs and decide which sound is the basic sound. Can any generalizations be made? (Hint Refer to natural classes.) a. [p!san] b. [dupykda] c. [pl}ko] d. [klsa] e. [SI}PWQ] f. [kamw:>kiti] 'full of leaves' 'bought' 'basket' 'we two' 'firewood' 'to move' g. [uduk] [kaskas) [poki) [pill rapid] I. [lud3uk) h. i. j. k. 'to 'to strike something' 'water' 'outrigger support' 'to tackle' Beginning Exercises 24. Sindhi The following data are fromSindhi,an Indo-European language of the IndO-Aryan family, spoken in Pakistalland Il1.dia. Examine the distribution of the pb.pnes [p), [ph), and [b]. Determine if the three are allophones of separate phonemes 0rallophones of the same phoneme. What is your evidence? Is the relationship among t he same as in English? Why or why not? a. [p:mu) [vad3u] [Jeki] [gado] b. c. d. e. f. [phanu] 'leaf' ,opportunity' 'suspicious' 'dull' 'door' 'hood of snake' g. [taru] h. [khato] i. [bad3u) j. [banu] k. [batJu) I. [d3ad3u] 'bottom' 'sour' 'run' 'forest' 'be safe' 'judge' 140 2 5. Standard Italian Consider the following data from Standard Italian, an Indo-European language of the Romance family, spoken in Italy. Answer the questions that follow. a. b. c. d. c. [tinta] [twda] [dantsa] [new] [d3w te] f. [sapone] 'dye' 'tent' 'dance' 'black' 'people' 'soap' g. [tiIJgo] h. [te1)go] i. [fu1)go] j. [bja1)ka] k. [a1)ke] I. [fa1)go] 'I dye' 'I keep' 'mushroom' 'white' 'also' 'mud' i. Are there any minimal pairs? If so, what are they, and what can you conclude to be true of Italian from those minimal pairs? ii. State the phonetic environments in which the sounds [n] and [1)] appear. Iden- tify any natural classes of sounds that appear in the enviwnments you've provided. iii. Given what you know about the distribution of sounds and the environments you listed in (ii), are [n] and [1)] in complementary or contrastive distribution? Please explain your answer. 26. Standard Spanish Standard Spanish is an Indo-European language of the Romance family. Examine the phones [d] and [6]. Determine whether they are allophones of one phoneme or of separate phonemes. If they are allophones of one phoneme, identify the o f distribution. If they are in complementary distribution, state a rule that describes the distribution. If [d] and [6] are allophones of separate phonemes, give minimal pairs that prove this. a. [drama] b. [dolor] c. [dime] d. [ka6a] c. f. 'drama' 'pain' 'tell me' 'each' 'side' 'hatred' g. [komi6a] h. randal i. j. k. I. [sweldo] [durar] [toldo] [falda] 'food' 'scram' 'salary' 'to last' 'curtain' 'skirt' 27. Russian Russian.is an Indo- European language of the Slavic family, spoken in Russia. Determine from following Russiandata whether [a} compleIneIlteach other as allophones of the same phoneme or whether they arein contrast as allophones of separate phonemes. If they are allophones of separate phonemes, provide evidence for your claim. If they are in complementary distribution, pick one allophone as the basic sound, and give the conditioning phonetic contexts for its allophones. ([t] represents a velarized [I], [sj] a alveolar fricative, and[m j] a voiced bilabial nasal.) c. [dar] d. [mas j] c. [mjata] 'atom' 'two' 'gift' 'ointment' 'mint' f. [upat] g. [dot] h. [partka] i. [ukrata] j. [brat] 'he fell' 'he gave' 'stick' 'she stole' 'he took' 28. Burmese Burmese is a Sino-Tibetan language of the Tibeto-Burman family, spoken in Myanmar. The following Burmese data contain both voiced and voiceless nasals. The latter are indicated by a small circle placed under the phonetic symbol. Are [m] and [IP] allophones File 3.6 Practice 141 of the same phoneme, or are they different phonemes? What about en] and [Q]? Is the same also true for (1)] and Lo]? Give evidence for your answer. If there is a phonological process involved, state what it is and give the conditioning environment. What is it about this environment that triggers this rule? Note: Burmese isa tone language, where indicates a high-toned vowel, ['] a low-toned vowel, a falling-toned voweL No tone marking indicates that the vowel is mid-toned. The sequence of sounds [eI] is a diphthong. n a. b. c. d. e. f. g. h. i. j. k. I. m. n [mil [mwel] [mjilJ [mjawn] [mjin] [ne) [nji?] [nwe] [hlpjaw?] [hQeI] [hQWeI] [hQja?] [hue?] 'fire' 'to give birth' 'river' 'ditch' 'to see' 'small' 'dirty' 'to bend fleXibly' 'to multiply' 'slow' 'to heat' 'to cut off (hair)' 'bird' n. o. p. q. r. s. t. u. v. w. x. y. z. [nie] [nwa] [l)a] [lJou?] [min] [hIpi] [hlpWeI] [hlpjaj?] [hlpown] [hQjilJ [hQjeI?] [hU a] Chin] 'fine, small' 'cow' 'five' 'stump (of tree)' 'old (people)' 'to lean against' 'fragrant' 'to cure (meat)' 'flour, powder' 'to wring, squeeze' 'to nod the head' 'to borrow' 'curry' 29. Korean Korean is a "language isolate," meaning that it is not linguistically related to other languages. It is spoken in Korea. In the follOWing Korean words, youwill find the sounds [s] and Determine whether the sounds [s] and [J] are allophones of the same phoneme or separate phonemes. If the sounds are allophones oUhe same phoneme, give the basic and derived allophones and the environment in which the derived allophone occurs. [n a. b. c. d. e. f. g. h. i. [fi] [miJln] [finmun] [th aksa1)j1ge] [filsu] [oJlp] [pa1)j1k] [kanJlk] [kaJl] 'poem' ,superstition' 'newspaper' 'table clock' 'mistake' 'fifty' 'method' 'snack' 'thorn' j. [sal] k. [kasu] I. [sanmun] m. [kas;)l] n. o. p. q. [miso] [susek] [tapsa] [so] 'flesh' 'singer' 'prose' 'hypothesis' 'smile' 'search' ,exploration' 'cow' 30. English English is an Indo-European language of the Germanic family. In the follOWing dialect of English, common in Canada and parts of the United States, there is a predictable variant [;)1] of the diphthong em]. What phonetic segments condition this change? What feature(s) characterize the class of conditioning segments? a. b. c. d. e. [b;)It] [tm] [lOId] [.Imz] [.I;)It] bite tie ride rise write f. [f;)lt] fight g. h. i. j. buy [bm] [.I;)IS] [fall] [1;)If] rice file k. [tmm] I. [t;)Ip] m . [name] n. [faq] o. [b'Hk] time type ninth fire bike 142 31. Totonac Examine the classes of voiced versus voiceless vowels in Totonac, a Totonacan language spoken in Mexico. Are voiced and voiceless vowels in Totonac in contrast, in free variation, or in complementary distribution? If the sounds are in complementary distribution, pick one sound as the basic sound and give the phonetic contexts for its allophones. (Note that [t5 ] represents a voiceless alveolar affricate, and [t] a velarized [1].) a. [t'aps\l-] b. [t'ilinks\l-] c. [kasittiJ d. [kukl}] e. [tkak\l-] f. [milq] 'he stacks' 'it resounded' 'cut it' 'uncle' 'peppery' 'snow' g. [snapap\l-] [stapl}] [fumpiJ [ta:qhl}] [tihaftiJ [tukftiJ h. i. j. k. I. 'white' 'beans' 'porcupine' 'you plunged' 'he rested' 'it broke' Intermediate Exercises 32. Tojolabal Tojolabal is a Mayan language of the Kanjobalan-Chujean family, spoken in Mexico. Determine whether plain [k] and glottalized [k?] are allophones of a single phoneme, in free variation, or in contrast. Support your answer with specific examples. (Hint: Don't forget that near-minimal pairs can be as convincing as minimal pairs.) a. b. c. d. e. [kisim] [t'ak?a] [koktit] [k'ak] [p?akan] f. [k?a?em] 'my beard' 'chop it down' 'our feet' 'flea' 'hanging' 'sugar cane' g. h. i. j. [sak] [k?ifin] [skutfu] [k?u:tes] k. [snika] I. [?ak?] 'white' 'warm' 'he is carrying it' 'to dress' 'he stirred it' 'read' 33. Spanish Examine the following data from Spanish and answer the questions which follow. Note that [{)] represents a voiced bilabial fricative, and [y] a voiced velar fricative. a. b. c. d. e. [bino] [di{)ino] [ka{)o] [su{)teraneo] [brotar] f. [imbjemo] g. [amiyo] 'he came' h. [u{)a] 'divine' i. [golpe] j. [gato] 'end' 'subterranean' k. [ayo] 'to sprout' I. [iya50] 'winter' m. [teIJgo] n. [leyal] 'friend' 'grape' 'a hit' 'cat' 'I do' 'liver' 'I have' 'legal' o. p. q. r. s. [siylo] [pwe{)lo] [la5fOn] [ka{)ra] [loyrar] 'century' 'village' 'thief' 'goat' 'to achieve' i. The allophones [b] and [()] are in complementary distribution, as are [g] and [y]. Determine the conditioning environments for each pair, and state a rule that describes the distribution of the allophones. ii. Refer to Exercise 26 (Standard Spanish) and the rule for the distribution of the allophones [d] and [5]. Describe the distribution of [b], [d], [g] and [()], [5], [y] in the most general terms possible, assuming each pair of allophones follows the same pattern. 34. Canadian French In the dialect of French (an Indo-European language of the Romance family) spoken in Canada, consider the distribution of [t] and [t'] (a voiceless alveolar affricate) in the data below. State their distribution and determine if they are allophones of one phoneme or of separate phonemes. [y] and [Y] are high, front, rounded vowels, tense and lax, respectively. F ile 3.6 Practice a. b. c. d. e. f. 'all' 'ended' 'such' 'stamp' 'timid' 'title' [tuJ [abut,J [tclJ [tebJ [tSimIdJ [fItJ 143 g. [telegramJ h. [tn::J i. [hlt"yrJ j. [minytJ k. [t"yJ I. [tSyb] 'telegram' 'very' 'culture' 'minute' 'you' 'tube' 35. German German is an Indo-European language of the Germanic family, spoken in Germany. Examine the voiceless velar fricative represented by [xJ and the voiceless palatal fricative represented by i n the German data below. Are the two sounds in complementary distribution or are they contrastive? If the sounds are allophones in complementary distribution, state the phonetic contexts for each allophone. a. b. c. d. e. f. 'eight' 'book' 'hole' 'high' 'flight' 'to laugh' [axtJ [bu:xJ [bxJ [ho:x] [fluxtJ [laxdnJ ' I' ' real' ' (he/she/it) would speak' ' to smile' ' to smell' ' to fence' g. h. i. j. k. I. 36. Farsi Farsi is an Indo-European language of the Indo-Iranian family, which is the most widely spoken language in Iran. In the following data, do [rJ, [r], and [rJ belong to one, two, or three different phonelIles? If they belong to different phonemes, give the pairs of forms which show this. If they are allophones of one (or two) phonemes, state the rule for their distribution. Which one would you choose to represent the phonemic form, and why? [rJ voiced trill a. b. c. d. e. f. [farsiJ [qcedriJ [rahJ [risJ [ruzJ [rJ voiced flap [rJ voiceless trill 'army' 'Persian' 'a little bit' 'road' 'beard' 'day' g. [aharJ h. [behtcerJ i. [hcerntowrJ j. [tIarJ k. [tJed3urJ I. [firJ 'starch' 'better' 'however' 'four' 'what kind' 'lion' m. n. o. p. q. [ahafiJ [bceradcerJ [befidJ [bifceIJgJ [tIera] r. [dafidJ 'starched' 'brother' 'go' 'pale' 'why' 'you have' 37. Bukusu Bukusu is a Niger-Congo language of the Bantu family, spoken in Kenya. The nasal prefix [n-J indicates that the verb is in the first person ('I eat, go, sing,' etc.). Two different processes occur when [nJ stands before another consonant. Look at these words and think about what is happening. The symbols [13J, [JlJ, and [xJ represent, respectively, a voiced bilabial fricative, a palatal nasal, and a voiceless velar fricative. a. b. c. d. e. f. g. h. i. [ndi:laJ [se:ndaJ [Jl d 3ll:IJgaJ [IJga13a] [mbi:maJ [xolaJ [mbuka] [tukaJ [tuna] 'I hold' 'I move' 'I watch' 'I divide' 'I weigh' 'I do' 'I perish' 'I cook' 'I break' j. [Jl d 3ina] k. I. m. n. o. p. q. r. [sunaJ [xala] [IJgeta] [ndasa] [mbulaJ [ndula] [fwa:ra] [mbalaJ 'I scream' 'I jump' 'I cut' 'I pour' 'I add' 'I roam' 'I trample' 'I dress' 'I count' (cont.) 144 i. How does the behavior of a nasal differ when it stands before the different types of obstruents (stops, fricatives, and affricates)? ii. There are two phonological processes at work here. What are they? iii. Write phonological rules to capture the facts about the nasal prefix In-I in Bukusu. Advanced Exercises 38. Greek Modern Greek is an Indo-European language spoken in Greece. Examine the sounds [x], [kl, a nd [c] in the following data. [k] represents a voiceless velar stop, [x] a voicea voiceless palatal fricative, and [c] a voiceless palatal stop. Which less velar fricative, of these sounds are in contrastive distribution, and which are in complementary distribution? State the distribution of the allophones. a. b. c. d. e. f. g. h. i. [kano] [xano] [cino] [kali] [xali] [ceri] 'do' 'lose' ' pour' 'move' 'charms' 'plight' 'eel' 'candle' ' hand' j. [kori] k. I. m. n. o. p. q. [xori] [xrima] [krima] [xufta] [kufeta] [cina] 'daughter' 'dances' 'money' 'shame' 'handful' 'bonbons' 'goose' 'china' 39. Ebira Examine the sounds [e] and [a] in the following data from Ebira, a Niger-Congo language of the Nupoid family, spoken in Nigeria. Do they appear to be allophones of separate phonemes or allophones of the same phoneme? If the two sounds are in complementary distribution, state the conditioning environments for the allophones. a. b. c. d. [mezi] [meze] [meto] [metu] 'I expect' 'I am well' 'I arrange' 'I beat' e. f. g. h. [mazI] [maze] [mat:)] [matu] 'I am in pain' 'I agree' 'I pick' 'I send' 40. Ukrainian Ukrainian is an Indo-European language of the Slavic family, spoken in Ukraine. Compare the masculine nominative singular forms of nouns with the vocative forms (nominative is used for the subject of a sentence, and vocative is used when calling to or addressing someone, as in "Hey, Robin."). There is a phonological change between the nominative and the vocative, which adds the ending [-e] to the nominative form. Three pairs of sounds are in allophonic variation. What are these pairs of sounds? What sort of phonological process is at work here? (There is a special name for it; see File 3.2.) What do you think is conditioning this alternation? (The symbols [h] and [x] stand for a voiced glottal fricative and a voiceless velar fricative, respectively.) Nominative a. b. c. d. e. f. g. Vocative Gloss [rak] [junak] [3uk] [pastux] [ptax] [bOh] [pIUh] [ratfe] [junatfe] [3utfe] [pastufe] [ptafe] [b03e] [pIU3e] 'lobster' 'youngman' 'beetle' 'shepherd' 'bird' 'God' 'plough' F ile 3.6 Practice 145 41. Maltese Maltese is an Afro-Asiatic language of the Semitic family, spoken on the island of Malta in the Mediterranean. Consider how the indefinite (a, some) and the definite (the) are formed in the following words. Maltese forms the definite of a noun by attaching either /iI-/ or /1-/ to it. Examine the data below and answer the questions that follow. (The symbol [h] represents a voiceless pharyngeal fricative.) a. Indefinite [fellus] [aria] [mara] [omm] [kelb] [{attus] [hitan] [abt] [ispaniol] 'chicken' 'air' 'woman' 'mother' 'dog' 'cat' 'walls' 'armpit' 'Spanish (language)' Definite [ilfellus] [laria] [ilmara] [lomm] [ilkelb] [il?attus] [ilhitan] [labt] [lispaniol] 'the chicken' 'the air' 'the woman' 'the mother' 'the dog' 'the cat' 'the walls' 'the armpit' 'the Spanish (language)' i. How can you predict the form of the definite marker? ii. What natural classes of sounds are involved? Now look at these nouns in the indefinite and the definite: b. Indefinite [tim] [dawl] [shab] [natura] 'a fig' 'a light' 'some clouds' 'nature' Definite [ittim] [iddawl] [isshab] [innatura] 'the fig' 'the Ught' 'the clouds' 'the nature' The definite marker has the same phonemic form in these words as it had in part (a), but a phonological process has changed its phonetic form. iii. What type of process is responsible for the change? How did it affect the definite marker? iv. What natural class of sounds causes the change from the phonemic form to the various phonetic forms in part (b)? v. Give the definite form of the following nouns: Indefinite [da:r] [zift] [azzar] [iI)gliz] [belt] 'a house' 'a pitch' 'a piece of steel' 'English' 'a city' Definite 'the house' 'the pitch' 'the steel' 'the English (lang.)' 'the city' ... View Full Document

End of Preview

Sign up now to access the rest of the document