The Nature of Processes
Modern phonological theory was invented by Jan Ignacy Baudouin de Courtenay (1845-1929), who studied the behavior of phonetic alternations in contemporary languages. Baudouin recognized two types of phonetic alternations: physiophonetic alternations, which represented a phonetic variation of a single phoneme, and psychophonetic alternations, which represented alternations between two distinct phonemes. The definition of that original dichotomy and of the phoneme itself have undergone many changes over the past century, but phonological theory is still very much grounded in the concept of alternating speech sounds. Modern generative theory tends to explain alternations in terms of phonological rules, but the approach to phonology is essentially alternation-based. I will address the relationship between Natural Phonology and Baudouin de Courtenay's theory later on. I only bring up Baudouin de Courtenay here to make a specific point--that you will not understand the Rule/Process dichotomy of Natural Phonology if you see it in terms of phonetic alternations. You have to start with a completely different conception of phonology and morphophonology. The Rule/Process dichotomy is grounded in psychological function, not phonetic alternations.Let us begin with the concept of a speech impediment. We all understand what a speech impediment is. It is an involuntary speech defect that causes speakers to mispronounce sounds in words. A person who lisps tends to replace "s" and "z" sounds with what sounds like a "th" sound. Basically, a phonological Process is a speech defect of precisely this sort, except that it is normally thought of as a "feature, not a bug" in normal adult pronunciation. That is, a phonological system, which consists entirely of Processes, is a system of constraints on pronunciation that cause speakers to mispronounce phonetic targets "correctly". Hence, processes differ from speech impediments only in that they are not always perceived as defects in articulation or mispronunciations. In normal adult speakers, they create expected, predictable pronunciations of intended phonetic targets. A Process is still a kind speech impediment, but it is normally an impediment to the mispronunciation of words in one's native dialect.
Let's consider two well-understood rules of phonology that correspond to Processes in Natural Phonology: aspiration of voiceless stops in English and syllable-final devoicing of obstruents in languages such as Russian. Aspiration and devoicing are speech impediments in the following sense. They cause speakers to mispronounce phonetic targets of a given linguistic dialect correctly and those of other dialects incorrectly. The English speaker who aspirates voiceless stops appropriately is not really perceived as mispronouncing those stops. Quite the contrary. But aspiration causes that speaker to mispronounce languages that do not permit aspirated voiceless stops. So the English speaker's tendency to aspirate the initial consonant in French père 'father' is perceived as a mispronunciation. Similarly, the Russian speaker who pronounces piterburg 'St. Petersburg' with a final unvoiced stop, is perceived as mispronouncing the word iceberg with a final "k" sound in English.
To summarize what I have said about Processes, they are psychological impediments to, or constraints on, the pronunciation of intended linguistic phonetic targets. They are only perceived as speech defects when they create unexpected misarticulations, a situation that normally does not occur in the pronunciation of one's native dialect. So, unlike conventional phonological rules, Processes are expected to apply cross-linguistically--to native and non-native pronunciations alike. They exist independently of intuitions of well-formed articulation, but they play a role in the mental calculation that yields intuitions of well-formedness, as I will explain in the next section. They are directly relevant to the study of linguistic articulation during speech performance. They are only indirectly relevant to the study of intuitions of linguistic well-formedness.
The concept of a "phonetic target" is important in an understanding of how Processes work. A "phonetic target" is theoretically an unconstrained pattern of sounds. Normally, phonetic targets are patterns or strings of juxtaposed sounds that are associated with the words of one's language. Sometimes, such phonetic targets are described as "phonemic". However, my concept of phonetic target refers to phonetic strings that represent words in a language-independent sense. One can attempt to pronounce any phonetic target, and it does not have to be considered a native word. If the target is thought of as lexical in principle, then Processes will constrain its articulation. If the target is just conceived of as random articulation, then Processes do not constrain it. In other words, Processes are part of a linguistic system, and they do not govern just any coordinated movements of the speech tract. This accounts for the well-known observation that people can mimic sounds that they find difficult to articulate as components of words.
Although I earlier compared two processes to conventional phonological rules, I emphatically deny that Processes are grounded in traditional notions of phonetic alternations. It is only a side effect of processes that they sometimes manifest themselves as alternations in linguistic data. In reality, some processes have no alternational footprint at all in a given language. They constrain the articulation of phonetic targets that simply fail to present themselves as a problem for the speakers of a given language. However, they do show up when speakers attempt to articulate foreign words or names that contain unfamiliar sequences of sounds, e.g. borrowed words, foreign personal names, and foreign place names.For example, the pronunciation of the Polish name Zbigniew presents an English speaker with the problem of articulating an initial [zb] cluster. Speakers tend to resolve that difficulty either by inserting an epenthetic vowel in the cluster or by devoicing the entire cluster, i.e. [z@bIgnu] or [spIgnu]. Both the epenthesis and devoicing are caused by Processes that do not normally give rise to observable alternations in English. They only manifest themselves under the unusual condition where a speaker tries to pronounce an unconventional phonetic target.
