Linguistics 220 Phonology: distributions and the concept of the phoneme John Alderete, Simon Fraser University
Foundations in phonology Outline 1. Intuitions about phonological structure 2. Contrastive and predictable distributions 3. The phoneme and conditioned allophones 4. A formal model: two levels of representation 2
Intuitions about phonological structure Same/different: s[t]op, cf. [tʰ]op English speakers will likely say same sound, but Thai or Mandarin speakers would say different Possible words: blick, *bnick; slish, *srish English speakers have intuitions about words they ve never heard before Hearing an accent: Ce n est [pʰ]as ca It s not that ; French speakers hear the aspirated word initial p as an English accent Ø Objective: provide a toolbox for describing these intuitions, and also giving a formal analysis 3
Sounds in contrast: phonemes Observation: in every language, there are pairs of sounds, or two sets of similar sounds, that contrast, i.e., they are the only sounds that signal a phonetic difference between two words Examples [p] vs. [b] [l] vs. [r] [i] vs. [ɛ] ta[p] ta[b] [l]eer [r]ear b[i]t b[ɛ]t [p]it [b]it [l]ip [r]ip b[i]d b[ɛ]d 4
Contrast, cont d Some terms Minimal pair: a pair of words that differ in exactly one phoneme Phoneme: a minimal contrastive sound Requirements for phoneme status Minimal: can t be other phonemes inside it Contrastive: must be able to signal a difference in meaning (i.e., there must be minimal pairs) Illustration Existence of pairs like pit and bit show that [p] and [b] are distinct phonemes. (Question: why?) Note: it s wrong to say that [p] and [pʰ] are distinct phonemes; there are no minimal pairs with these sounds, so they are not contrastive. Also: it s wrong to say that [pɪ] and [bɪ] are distinct phonemes; they are not minimal. 5
Contrast, cont d Task: provide the necessary evidence to show that the pairs of sounds below are in fact phonemes. [t] and [d]: [ɪ] and [ɑ]: [θ] and [ð]: 6
Predictable distribution: allophones Observation: in every language, there are phonetically different sounds that are not contrastive; instead they have a predictable distribution, i.e., the phonetic environments they occur in are systematic and regular. Illustration [p] and [pʰ] are different sounds, but they are not contrastive. [pʰ] occurs in the beginning of stressed syllables, and [p] occurs in all other positions. 7
Predictable distributions, cont d Some terms Complementary distribution: a state in which two or more sounds occur in non- overlapping environments; = sounds that never occur in the same positions Allophones: variants of the same phoneme whose distribution is predictable Illustration [p] and [pʰ] are in complementary distribution. [p] and [pʰ] are allophones of the same phoneme. (Explain why.) 8
Different representations Assumption We assume that phonemes and conditioned allophones are actually different kinds of phonological representations. Typography: / / (phonemes) versus [ ] (allophones). Example Phonemic representation (phonemes): /p/ Phonetic representation (allophones): [pʰ] [p] Implications: Native speakers typically distinguish different phonemes as different sounds, but not so of allophones. Native speakers can also detect an accent as an allophone that is out of its normal set of environments Sound inventories of individual languages need to distinguish phonemic inventories from phonetic inventories, where phonemic inventories are a proper subset of phonetic inventories 9
Phonemes are language particular Observation: the phonemic status of two sounds is a fact about a particular language; if two sounds are phonemes in one language, that does not mean they are phonemes in another language, and vice versa. Examples [ɛ] and [æ] are different phonemes in English, but conditioned allophones in Turkish, as shown by different realizations of the 1 st singular I, [bɛn] ~ [bæn] Short and long vowels are not contrastive in English, but many languages have distinct phonemes that differ only in vowel length, as in Japanese [tori] bird, cf. [tori:] shrine gate 10
Importance of natural classes, part I Observation: sometimes two sets of sounds show the same pattern of complementary distribution; the sets are corresponding pairs of conditioned allophones; these sets typically fall into natural classes of phonetically similar sounds Generalization: voiceless stops are aspirated the beginning of syllables; their unaspirated counterparts occur everywhere else. Illustration Phonemes /p/ /t/ /k/ Allophones [pʰ] [p] [tʰ] [t] [kʰ] [k] Environ- ments syllable initially elsewhere syllable initially elsewhere syllable initially elsewhere 11
Importance of natural classes II Observation: the environments for describing the distribution of conditioned allophones often depend on natural classes of sounds Generalization: voiceless [l ] occurs after voiceless stops, [l] occurs elsewhere Illustration Before voiceless stops Elsewhere plow [pl aw] blue [blu] clap [cl æp] Lou [lu] play [pl ej] slip [slɪp] 12
Exercise Task: what do the phonetic facts at other liquids, and indeed glides, like the examples shown below, indicate about the natural classes involved in post- stop voicelessness? Examples: brew, green, cf., prow, trip, creep Beauty, Duane cf., putrid, twin, quick, cute Revised generalization: 13
Summary of method used in phonemic analysis When considering the distribution of a set of sounds: First, test for distinct phonemes by trying to find minimal pairs. Second, if inconclusive, test for allophones of the same phoneme by trying to find a conditioning environment for one or more sounds Step 1: list all environments of both sounds Step 2: ask if there is any overlap in the contexts Yes à not allophones; sound difference is not predictable; look harder for minimal pairs No à Can you characterize one set of environments in simple phonetic terms? If yes, the distribution of the sound occuring in a well- defined environment is predictable and the two sounds are allophones of the same phoneme. 14
Exercise: [k] and [k ] in Tojolabal (Mexico) Task: determine whether [k] and [k ] are distinct phonemes or allophones of a single phoneme. kisim my beard tʃak'a chop it down koktik our feet k'ak flea p'akan hanging k'aʔem sugar cane sak white k'iʃin warm skutʃu he is carrying it k'u:tes to dress snika he stirred it ʔak reed 15
Tojolabal, cont d [k] environments [k ] environments 16
Tojolabal, cont d Interim conclusion: Additional evidence: [k ak] flea, cf. [k ak ] my grass Conclusion: 17
[t] and [tʰ] in Tojolabal Task: determine whether [t] and [tʰ] are distinct phonemes or allophones of the same phoneme. tʃitam pig tʃatatʰ another kind of plant makton a patch mutʰ chicken potot kind of plant nahatʰ long tinan upside down ʔinatʰ seed 18
Some take home messages Describing distributions: phonetic science gives us some tools for describing the patterns of sounds in utterances, including contrastive and predictable distributions. Two representations: we distinguish phonemes and phonetic sounds with two levels of representation. Phonemes are the contrastive sounds of a language, and their conditioned variants are shown in a different level. 19
A parallel between phonology and morphology Observation: natural classes, as described in phonetic terms, are relevant to predicting both phonological and morphological forms. The parallel: Phonetically conditioned allophones: different variants of the same phoneme, predictable from the phonetic environment Phonetically conditioned allomorphs: different variants of the same morpheme, predictable from the phonetic environment 20
Two birds with one stone Allophones: predict one or more conditioned sounds on the basis of phonetic environment [pʰ] /p/ [p] Allomorphs: predict one or more conditioned morpheme variants on the basis of phonetic environment /- z[plural]/ [- z] [- s] [- ɪz] 21
One stone: phonological rules Assumption: just like other linguistic processes, there are a set of phonological processes, with input conditions and output changes Example: Aspiration rule Input condition: a voiceless stop in the beginning of a syllable Output change: change the voiceless stop to its corresponding aspirated stop. 22
Phonemes and allophonic rules Assumption: to account for the two kinds of distributions (contrastive and complementary), we assume that there are set of basic phonemes in the input ( underlying representation ), and one or more rules that changes phonemes to their phonetic representation. Example Word bit pit spit Input /bɪt/ /pɪt/ /spɪt/ Aspiration rule does not apply pʰɪt does not apply Output [bɪt] [pʰɪt] [spɪt] 23
A subtle point: input restrictions Assumption: for the model to account for complementary distribution, it is crucial to state restrictions on the input; the outputs of allophonic rules never occur in inputs. In essence, the inputs are only made up of the basic phonemes. Task: show why this is necessary by considering a model where [pʰ] can occur in inputs. Are [p] and [pʰ] still in complementary distribution? 24
Question: how determine which sounds are basic phonemes? Answer: it is a matter of analysis. Just like with phonetically conditioned allomorphs, the analyst must consider all possible analyses; analytical reasoning supports one hypothesis over others. Task: consider the alternative hypothesis that [pʰ] is the basic phoneme, and [p] is the conditioned allophone. What rule is required, and is it more elegant that the alternative aspiration rule? In your analysis, consider words such as pit, spit, sip. 25
Occam s Razor O.R.: All other things being equal, the simplest solution is the best, translation: if the data does not require more complex assumptions, select an analysis that makes a few simple assumptions. Task: how does Occam s Razor apply to the assumption of [p] or [pʰ] in the input? 26
Approach to allomorphy Assumption: suppletion aside, morphemes have a single form in the input; but they may be changed in their phonetic representation by phonological rules. The specific consonants or vowels that change in allomorphy are called alternations. Example: Devoicing Input condition: voiceless sound + voiced sound at the end of a word Output change: changed the voiced sound to its voiceless counterpart Word dogs cats Inputs /dɑg + z/ /kæt + z/ Devoicing does not apply kæt- s Outputs [dɑgz] [kæts] 27
Further evidence for phonological rules: child language Assumption: children may have phonemic representations resembling adult words, but their production of these forms lags behind their perception, resulting in systematic changes, as illustrated below for a young child. These can be accounted for with phonological rules. Task: what generalizations govern the differences between the adult and child outputs? Account for it with one or more phonological rules. Word clean draw please Transcription kin da piz friend fɛn 28
Summary of results Contrastive phonemes: we represent the contrasts of a language by allowing them to be listed in the phonemic representation, which corresponds to the input to a phonological grammar. Conditioned allophones: the complementary distribution of phoneme variants follows from assuming that all but the basic phoneme are the result of phonological rules Conditioned allomorphs: the same phonological rules needed for allophones can also account for the systematic sound changes observed in allomorphy. Further evidence: child language 29