Title: Natural phonology and loanword phonology (with selected examples from Miami Cuban Spanish)
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00098665/00001
 Material Information
Title: Natural phonology and loanword phonology (with selected examples from Miami Cuban Spanish)
Physical Description: x, 429 leaves : ill. ; 28 cm.
Language: English
Creator: Bjarkman, Peter C
Copyright Date: 1976
Subject: Grammar, Comparative and general -- Phonology   ( lcsh )
Spanish language -- Dialects -- Florida -- Miami   ( lcsh )
Linguistics thesis Ph. D   ( lcsh )
Dissertations, Academic -- Linguistics -- UF   ( lcsh )
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Statement of Responsibility: by Peter Christian Bjarkman.
Thesis: Thesis--University of Florida.
Bibliography: Bibliography: leaves 416-427.
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00098665
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000178676
oclc - 03125271
notis - AAU5189


This item has the following downloads:

naturalphonology00bjar ( PDF )

Full Text






Peter C. Bjarkman


This dissertation is for MARY ANITA who was its inspiration and BOHDAN

who in the end made it possible.


Besides all the expected types of indebtedness surrounding any

advanced graduate education, this dissertation in one special sense or

another owes some small part of its existence to each of the following

institutions and individuals:

My advisors, Bohdan Saciuk and Gary Miller, for remaining always

patient and providing so many inspirations along the way. Much of

what I know about phonology can be credited directly to Gary Miller;

and Bohdan Saciuk has been the kind of teacher, counselor, and personal

friend I am not likely, soon if ever, to meet again.

The Graduate Council and the Faculty of the Program in Linguistics,

for generously awarding the Graduate Council Fellowship and two Graduate

Research Assistantships which have made my studies at the University

of Florida financially possible.

The Department of English at George Mason University for a faculty

appointment which provided the needed final impetus for my work.

David Stampe, for his early encouragement and for providing me

with a topic of relevance. Also, for better or for worse, David's

unbroken silence has repeatedly forced me to reach my own unique and

hopefully fresh solutions to the problems and issues his theories have

inevitably generated.

Julie Lovins, for her painstaking and encouraging comments, her

genuine interest, and several letters from "the land of the rising sun."

Richard Rhodes, for his many enthusiastic and pregnant letters

and for the motivations for Chapter Two.

Tracy Terrell, for responding rapidly and generously to my

requests for his own papers and works-in-progress on Cuban Spanish --

these works having provided an inspiration as well as a practical and

indispensable guide.

Bill Carrillo and Sergio Arnaiz, for all their aid in tracking

down several Miami informants.

Joan Hooper, Jorge Guitart, Joseph Matluck, Richard Wojcik, and

Jonathan Kaye, for answering all my demanding and sometimes impertinent

letters and requests.

Mrs. John B. Whitworth, my mother-in-law, for financial and

emotional support when it was most needed and for the gift of a tape


Mr. and Mrs. Robert Bjarkman, my father and mother, for so many

intangible gifts over the years which have helped in some small measure

to make me a better scholar and a better teacher.

And Mary Anita, who after everything is said and done is, above all,

the one who made it possible.

Somewhat belatedly, I would also like to thank Chauncey Ch'u,

whose classes were an enjoyable and satisfying learning experience,

whose own time was never too precious to spend a few moments with his

inquisitive students, and whose friendship I will treasure for a long

time to come as I look back fondly on my Gainesville experience.




Part One Introduction to Natural Phonology

1.1 Introduction 2
1.2 Claims of Natural Phonology 10
1.2.1 The innateness of natural processes 11
1.2.2 The restrictive nature of phonological
processes 12
1.2.3 The revisions of the innate system 13
1.2.4 The natural order of acquisition 14
1.2.5 The internal representations of adult speech 15
1.2.6 The innate residue in adult grammars 16
1.2.7 Naturalness motivations in linguistic change 18
1.2.8 The prominence of natural loanword phonology 20
1.3 Explanation and Performance Models 25
1.4 Remarks on Cuban Spanish and Methodology 40

Part Two The Processes of Natural Phonology

2.1 Preliminary Remarks 57
2.2 Conceptions of Rules and Natural Processes 60
2.2.1 Non-suspendability of natural processes 81
2.2.2 Optionality of natural processes 82
2.2.3 Extremity of natural processes 83
2.2.4 Non-linearity of natural processes 84
2.2.5 Phonetic functioning of natural processes 99
2.2.6 Allophonic influences of natural processes 100
2.3 On Paradigmatic and Syntagmatic Processes 113


3.1 Explanation in Stampean Natural Phonology 146
3.1.1 Hypotheses in Natural Phonology 150
3.1.2 The uniqueness of Natural Phonology 176
3.1.3 Two cases of reanalysis 188
3.2 Naturalness and Some Notational Variants 196


4.1 Surface and Underlying Constraints 209
4.1.1 The use of morpheme structure conditions 215
4.1.2 The use of surface phonetic constraints 231
4.2 Active and Passive Grammatical Components 240
4.3 On the Reality of Morpholexical Rules 252

Part Three The Applications of Natural Phonology

5.1 Introduction 264
5.2 Issues in Cuban Phonological Studies 265
5.3 Some Selected Processes in Miami Spanish 286
5.4 Implications of Natural Processes 290


6.1 Structural Implications of Borrowing 305
6.2 Earlier Inadequate Approaches to Borrowing 310
6.2.1 Haugen and Weinreich and Structuralism 311
6.2.2 Chomsky and Halle's Admissibility Theory 314
6.2.3 The "Magnetic Attraction" Hypothesis 317
6.3 The Applications of Natural Loanword
Phonology 321


7.1 Extent and Types of Borrowing 333
7.2 Phonological Nativization and Phonological
Adoption 348
7.3 Conclusions and Revisions in Natural
Loanword Phonology 350

Part Four Syllabic Phonology

8.1 A Theory of Syllabic Phonology 365
8.2 Consonantal Strength Hierarchies 369
8.3 Weakening Chains in Natural Phonology 387
8.4 Strength and Spanish Syllable Structure 393




Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy



Peter Christian Bjarkman

December, 1976

Chairman: Bohdan Saciuk
Major Department: Linguistics

This dissertation provides inter alia the first extensive application

of Natural Phonology proposed by Stampe to an investigation of linguistic

borrowing. My predominant purpose is to test out some central claims

of Natural Phonology -- a nascent explanatory theory of phonological

components in the grammars of natural languages being developed by

Stampe and his students from the initial assumption that a phonological

system is largely the residue of an "innate" system of phonological

processes which are significantly revised by the early activities of

language learning. It is demonstrated with these eight chapters that

claims made by the natural phonologists draw considerable support, among

elsewhere, from findings in loanword phonology: loanwords offer examples

corroborating the observations about child language acquisition which

so far make up the bulk of Stampe's evidence. It is also the purpose

of this work, however, to suggest revisions and amendments at points

where Stampe's theory has already proven intractable and insufficient

in the face of a large corpus of cross-linguistic data.


Part One is a comprehensive introduction to Natural Phonology.

The three chapters of Part Two provide a theoretical background by

evaluating some strengths and potential weaknesses of Stampe's pioneer

model. Chapter Two clarifies several misconceptions existing in the

current literature concerning "rules" versus "natural processes," and

on the status of the natural phonemic level of representation which is

the relevant level of abstraction in Natural Phonology. Chapter Three

outlines some additional hypotheses which are the foundations of Natural

Phonology -- the Innateness Hypothesis, the Suppression Hypothesis, the

Acquisition Hypothesis, the Phonemic Hypothesis, the Phonetic Change

Hypothesis -- and summarizes several supporting arguments which have

appeared only in Stampe's unpublished work (circulating preliminary

papers) and are likely unfamiliar to many linguists. Chapter Three

also takes up differences between Stampe's version and often incongruent

competing variations on Natural Phonology and examines the related

problematical areas. In Chapter Four a revised Stampean model is

suggested as part and parcel of a more comprehensive Natural Process

Morphophonology, incorporating within it a separate word formation or

morphological component. This revised model is intended to make lucid

precisely how Natural Phonology might more adequately account for both

underlying and surface level constraints, which together operate on

both the native and "exceptional" (foreign) phonological forms.

Discussion in Part Three of this study is devoted to the native

phonology and loanword phonology of Miami Cuban Spanish as these are

revealed through apparatus of Natural Phonology. Whereas Part Two is

concerned largely with defining the concept and role of processes in

Natural Phonology, Part Thred emphasizes actual applications of the


theory. More specifically, Part Three explores the implications of

principles of Natural Phonology for the analysis and explanation of

loanword data and in the process supports and illustrates claims of

the original Stampean model of Natural Phonology as well as the revised

model rendered in Chapter Four. This integral part of the dissertation

(especially Sections 7.1 through 7.3) extends a series of claims made

by Julie Lovins (1973 University of Chicago Dissertation) that loanword

phonology essentially is natural phonology and that phonological systems

are perhaps actually best revealed through active processes in linguis-

tic borrowing. Evidence for the dominant role of natural processes in

loanword assimilation derives from tapes of bilingual as well as nearly

monolingual speakers of the Miami Cuban dialect of Spanish, a dialect

which is an unparalleled source of contact borrowing in Spanish phono-

logical studies.

Finally, Part Four demonstrates how fundamental theses of Natural

Phonology will conspire along with dialect-specific examples from

borrowing to modify certain basic assumptions in the literature about

the role of syllable structure in an explanatory phonological grammar.

Part One

Introduction to Natural Phonology



Natural Phonology is essentially a continuous search for the op-
timal grammar which is consistent not only with language-specific
data but also with known properties of human languages in general.
Matthew Y. Chen

1.1 Introduction

A primary justification for this present dissertation is the

paucity of previous conclusive work on loanword phonology, while another

is the incomplete status and relative unavailability of David Stampe's

own pioneer efforts in developing the model of Natural Phonology. With

his few articles (Stampe 1969, 1972, 1973c) and circulating papers

(Stampe 1968, 1973b), Stampe provides limited examples from outside of

child language and fewer still from outside English. And though it is

his own assumption (see especially Stampe 1969, page 451, footnote 4)

that loanword phonology provides the crucial testing grounds for any

phonetically-based and/or phonologically-based model, very little has

yet been said by Stampe himself or others about loanwords of specific

dialects in any of the available accounts of Natural Phonology.1

The theory of Natural Phonology which Stampe (1969) proposes

as an essential modification of Chomskyan "generative" phonology is

based almost exclusively on certain empirically testable hypotheses

about the universal physiological basis of principles governing language

acquisition and language change. There is the hypothesis, for example,

that the phonological system of a language is an innate system of


physiologically-induced articulatory processes, as well as the connected

hypothesis that language acquisition is largely a matter of randomly

suppressing these innate universal processes.

These hypotheses are first outlined in Stampe's earliest publication

(Stampe 1969) and more recently expanded with finer illustration in his

seminal doctoral dissertation (Stampe 1973a). They have become as well

the cornerstones for publications on both child language acquisition and

linguistic borrowing authored by a small force of Stampe's students and

others prominently supporting Natural Phonology (viz. Lovins, Edwards,

Ohpo, Wojcik, Bjarkman). Yet it still remains that the literature on

Natural Phonology is not at all extensive. An almost complete listing

of the works devoted to advocating Stampe's theories would include no

more, perhaps, than the following dozen or so entries: Darden 1971;

Edwards 1970; Hutcheson 1973; Lovins 1973, 1974a, 1974b; Donegan Miller

1972a, 1972b, 1973a, 1973b; Nessly 1973; Ohso 1971; Rhodes 1972, 1973,

1974; and most recently Wojcik 1976. Also there are earliest drafts of

sections of this dissertation, which have appeared as Bjarkman 1974a,

1974b, 1975, 1976. Articles or monographs most severely critical of

Stampe's claims, on the other hand, would especially be the following:

Dressler 1974; Hooper 1975; Lee and Howard 1974; Miller 1975; Ohala

1974; and a dissertation by Stoel (1974). Of these latter works, those

by Dressler, Lee and Howard, and Ohala have all simultaneously appeared

in the single Chicago Linguistic Society volume entitled Papers from

the Parasession on Natural Phonology (cf. Bruck et al. 1974).

There is, of course, a plethora of brief references to Stampe's

theories throughout numerous papers and monographs on child language

(e.g. Ingram 1974 and Golick 1974). But it is not at all inappropriate

to observe at this point that exceptionally little is offered within

this second list of papers to supplant or even noticeably weaken crucial

assunptions of the natural phonologists. To date, the theory has been

superficially praised in some quarters while being altogether ignored

in others.2 But Natural Phonology has surprisingly not been very

seriously challenged even by its severest critics, nor has it yet been

very adequately tested out by even its most enthusiastic proponents.

This dissertation will undertake answering some of the largely

superficial objections to Stampe's theory, as well as most of the more

substantive criticisms of Natural Phonology. I will both support the

model where it offers elegant solutions to long-standing problems in

phonological theory and at the same time offer several amendments where

there are inevitable shortcomings in its present version. My approach

is one designed to combine a double purpose: (1) of explicating Natural

Phonology in some of its more elusive aspects and (2) of exploring its

pragmatic applications to what has proven to be a sparsely attended

field of loanword theory.

The four chapters in Parts One and Two provide a sketch and an

expansion of the present model of Natural Phonology. Chapter Two,

especially, clarifies Stampe's distinctions between the basic innate

"natural processes" (i.e. those substitutions which are mentalistic

and physiological in their motivations) and the acquired "rules"

(i.e. the standard-type phonological rules of traditional generative

phonology) and establishes a working conception of the "natural phonemic

level" postulated for Stampe's model. Some theoretical arguments

against a level of phonemic analysis and also against Stampe's pro-

posal of paradigmatic or context-free natural processes are dismissed


as being without sufficient foundation. Chapter Three takes up further

evidence for the Phonemic Hypothesis and other crucial hypotheses of

Natural Phonology. A distinction is made between the version of Natural

Phonology proposed by Stampe and other notational variants, such as

Vennemann's Natural Generative Phonology (Vennemann 1971, Hooper 1973)

and Schane's notions of naturalness (Schane 1973). An overview is also

presented of some troublesome inconsistencies in the theory as Stampe

formulates it to date. Stampean phonemics are compared to classical

phonemics and it is claimed that a level of phonemic representation

is not only motivated empirically but also capable of withstanding

effectively each of Halle's and Chomsky's famous arguments against the

phoneme as a viable linguistic concept. These first several chapters

are, then, a necessary summary and explication of the evidence and

assumptions underlying a more or less still tentative model for Natural


Chapter Four amends Stampe's view of a phonological component to

allow additionally the rules of morphology apparently needed to explain

some of the evidence from loanword phonology as well as to account for

other notable deficiencies in current generative theories. The type of

separate morphological component proposed is similar to one originally

envisioned by Halle (see also Aronoff 1976 and Harris 1974) with a

list of morphemes, rules of word formation, and a lexical exception

filter. It suggests as well a necessary division between an Active

Grammar, which comprises the syntactic and phonological components

through which all new derivations automatically pass, and a Passive

Grammar, which is the paradigmatic processes and morphological component

which are only irregularly operable in the adult version of the language


My contention in Chapter Four is also that recent arguments for

surface phonetic constraints (the SPC's) as an alternative to morpheme

structure conditions (the MSC's) in generative grammar represent a

false dichotomy between actual surface phonotactics and the underlying

morphophonemic rules. That such underlying constraints on "morpheme"

structure do not now seem to be a valid theoretical consideration

(such restrictions in Spanish and apparently in most languages are

on "words" not "morphemes") does not altogether preclude "lexical"

as well as "phonetic" constraints in a generative grammar. It is

proposed in this chapter that what have traditionally been referred

to as SPC's are captured within a revised Stampean model in terms

of active applications of the syntagmatic natural processes and are

not at all some kind of "static conditions" in the grammar. Similarly,

underlying restrictions (formerly the MSC's) must be taken not strictly

as Stampe proposes (as elusive paradigmatic processes which supposedly

are a type of filter on the lexicon: i.e. the process, for example,

which eliminates all nasalized vowels from the lexicon of English or

the one blocking underlying velar nasals in this same language), but

rather as a concert of syntagmatic processes acting in conjunction

with the templates which are the word formation rules of what we will

call the "passive" morphological component.

It is shown, finally, that recent proposals in Sommerstein 1974,

Shibatani 1973, Clayton 1976, and elsewhere for phonotactically motiv-

ated rules as a modification of the standard generative theory are valid

only if the questionable assumption is made that all phonological sub-

stitutions are of an identical teleology. Natural Phonology, with its


basic distinction between innate processes and acquired rules, of

course denies this overly strong assumption.

Discussion in Part Three is devoted to the native phonology and

loanword phonology of Miami Cuban Spanish, as these are revealed through

apparatus of Natural Phonology. Data is drawn from quality cassette

tapes of twelve primary informants, obtained in formal or semi-formal

interview sessions, as well as from fifteen supplemental tapes (about

ten hours) of Spanish radio programming and unstructured "free-speech"


Chapter Five examines identifying characteristics of the Miami Cuban

dialect, e.g. random Velar Nasal Intrusion (velarization of syllable-

final nasals), voicing of voiceless intervocalic stops, spreading of

the related obstruent spirantizing process (surfacing of voiced stops

as spirants), and final consonant deletions. It is demonstrated, for

example, that the rampant phenomenon of Velar Nasal Intrusion by which

most final nasals are realized in systematic fashion as velars is a

progressive-type assimilation involving preceding vowels and hence a

process of general weakening related to most other weakening processes

characteristic of this dialect. The argument is developed that evidence

from loanword phonology and principles of Natural Phonology combine to

suggest also that predominance of spirantized voiced obstruents results

from the spreading of a natural process characteristic of an "easy-

articulation" dialect and not from any phonemic restructuring of under-

lying forms (as implied in Hammond 1975, 1976). Chapter Six, in turn,

assesses past failures to deal insightfully with data from linguistic

borrowing under a variety of theoretical approaches and suggests, by

contrast, the full explanatory powers characterizing what Lovins (1974b)

has labelled Natural Loanword Phonology.

Chapter Seven is a central chapter which extends the explanatory

powers of Natural Loanword Phonology to an examination of the precise

role of processes, opposed to learned rules, in shaping perceptions and

assimilations of foreign segments. The goal of this chapter is to verify

with representative data from Miami Spanish some following types of

assumptions about Natural Loanword Phonology.

Hypothesis I: A language's ability to borrow words is to a major

extent determined by what shapes it permits at the surface level; that

is to say, it is generally the late syntagmatic processes and not the

underlying restrictions paradigmaticc processes or morphological rules)

which are relevant in determining what can be borrowed and which operate

on what lexical items eventually are borrowed into a target language.

Hypothesis II: Loanwords have underlying representations first

established at a depth corresponding to Stampe's notion of the natural

phonemic level as a level of linguistic reality.

Hypothesis III: Only processes, which are "alive" and productive

in the grammar, and not rules, which remain "dead" and unproductive,

will function in the initial establishing of a loanword in the target


Hypothesis IV: Lexical forms of the target language are definable

as either native forms (those to which all "rules" have applied wherever

applicable); assimilated loans (those to which all available processes

have applied); and unassimilated loans (those which contain violations

of processes constraining the underlying representation and are there-

fore "phonological" exceptions, and those which contain violations of

processes determining surface representation and are therefore "phonetic"


Hypothesis V: Substitution processes which will not serve to

predict structures of elements entering the target language are easily

classifiable as rules rather than processes, and this is, in fact, one

foolproof measure of what is a "rule" and what is a "process" for any

natural language.

These assumptions are tested with the treatment of abundant English

loanwords adopted by Miami Cuban speakers of Spanish. One conclusion

which seems incontrovertible in light of data available is that moribund

"rules" are clearly irrelevant to initial lexicalization of loanwords

in the target language, though they do likely afterwards apply at some

post-borrowing stage to bring about greater "nativization" of foreign


In Part Four a developing theory of Syllabic Phonology is discussed

in relation to both the proposals of Natural Phonology and the demands

of loanword phonology. This theory is one based on the postulation of

"weakening chains" in Natural Phonology (Nessly 1973) and modified along

lines suggested by Hooper (1973) relative to positive syllable structure

conditions for Spanish, which in turn build upon a notion of universal

consonantal strength hierarchies. The object of study here is speakers'

apparent strategies for syllable structure modification if borrowing

takes place from some language with complex syllable structure (English)

to one with less complex syllable structure (Spanish). Attempts are

made to demonstrate quite irrefutably that the modifications of syllable

structure in at least this one dialect are more insightfully explained

when assumptions of the Stampean natural phonologist are adopted as a

standard. Heavily syllable-based schemes such as Hooper's version of

Natural Generative Phonology are, on the other hand, neither explanatory

nor even consistently accurate in their predictions.

In brief, this dissertation establishes why loanword phonology is

central to progress in phonological theory in general. This primacy of

loanword phonology stems, of course, from the fact that the structure

of a lexicon is product of the workings of a phonological system whose

functions and whose mechanisms are nowhere better illustrated and more

available for study than in the processes of linguistic borrowing.

This present study is motivated throughout, then, by an assumption

that Natural Phonology can best approach precisely those problems raised

by linguistic borrowing which have remained insoluable within several

previous frameworks for grammatical study. One reason seemingly is

the explicit distinctions set forth in this theory between processes

and acquired rules. Another is the focus on casual speech phenomena.

A third is the positing in Natural Phonology of quite non-abstract

phonetically and mentalistically motivated underlying representations

for all surface strings. But the primary advantage of Natural Phonology

over generative phonology or structuralist phonology or stratificational

phonology emerges from the fact that the latter undisputably are all

descriptive enterprises while only the former provides us with a truly

explanatory model of linguistic behavior.

1.2 Claims of Natural Phonology

Stampe's hypotheses governing the acquisition and development of

phonological systems may be briefly and somewhat superficially outlined

for this chapter after the following fashion, occasionally paraphrasing

here from Stampe's own discussions in Stampe 1969 and elsewhere.


1.2.1 The innateness of phonological processes

A phonological system for any speaker of a.,y language is the

residue of an innate system of phonological substitution processes,

already possessed in its entirety by the infant first encountering his

language and subsequently revised extensively during the course of the

child's early acquisition of his native speech.3 Stampe defines a

phonological "process" as

a mental operation that applies in speech to substitute,
for a class of sounds or sound sequences presenting a specific
common difficulty to the speech capacity of the individual,
an alternative class identical but lacking the difficult
property (1973a, page 1).

This is a way of saying that processes merge potential opposition

into members which present the least obstacle to the speech-producing

mechanisms. To illustrate: in an environment favoring voicing (such as

intervocalic position) such a merger would voice all obstruents, which

by nature and irrespective of context are, in the absence of such pro-

cesses, more easily articulated as voiceless. The pressure to devoice

obstruents (i.e. to produce them as voiceless) irrespective of context

-- their oral constriction being an impediment to an airflow needed

for voicing -- but yet to voice them between vowels demonstrates that

processes by nature come in contradictory sets revealing conflicting

phonetic restrictions. When such conflicting tendencies arise, Stampe

assumes the speaker will resolve them by means of three observable

mechanisms: (1) he suppresses one of the set of contradictory pro-

cesses; or (2) he places some limitation on the set of segments to

which a process will apply or the context in which it applies, and

thus he suppresses some part of the process; or (3) he orders the

application of the processes to achieve a particular output. From

babbling to the achievement of adult pronunciations, the suppression,

limitation, and ordering of procLes- results in the considerable re-

vision of the child's innate system of phonological substitutions.

1.2.2 The restrictive nature of phonological processes

The inherited innate phonological system represents a complete

"potential" system of phonological as well as mental restrictions on

speech, and these arc manifested in a type of process which makes

real substitutions, even though this type clearly differs ontologically

and teleologically from the learned "rules" in the phonology. Stampean

processes are restrictions on phonological output, but they are not

static conditions in the sense of the templates which are strictures

on morpheme structure (Halle 1973) or in the sense of redundancy rules

as these are conceived in the standard theory (Stanley 1967).

We must speak here of a "potential" system, for despite a reason-

able similarity in approach, different children learning the same lang-

uage will display different sets of processes, and in a variety of pos-

sible ordering (Stampe 1973a; Ingram 1974; Salus and Salus 1974). Also,

the inventory of processes relevant to a single language (or a single

speaker of a single language) is only a portion of the total inventory

available to each new human speaker.

With the onset of the child's speech in the babbling stage the

natural processes appear to be entirely unordered, and the most extreme

processes are, in fact, directly observable only at this primitive

stage (Stampe 1969, page 444). The earliest substitutions of the >

child's phonology are only these more extreme processes: e.g. un-

stressed (weak) syllable deletion, cluster reduction, obstruents

becoming lax stops, merger of vowels to single central articulations


(the maximally vocalic a) (Stampe 1969; Salus and Salus 1974). From

the outset the child's speech will display both syntagmatic processes,

which function to ease transitions from one segment to another, and

paradigmatic processes, which establish properties of individual

segments, though the preponderance of earliest processes is in the

direction of the paradigmatic. Only with young children do the para-

digmatic processes typically apply in "feeding" orders, and that this

should be so is a natural outcome of the fact that adult speech would

find intolerable the massive kinds of neutralizations of underlying

distinctions which mark infant speech and which several paradigmatic

processes applying in a feeding relationship would work to produce

(Donegan Miller 1973b).

1.2.3 The revisions of the innate system

The acquisition of new phonetic opposition (i.e. new single

segments) by the child entails a constant revision of innate processes

of his phonological system. We have already noted Stampe's central

claim that the mechanisms for resolving conflicting processes are the

same mechanisms relied upon for revising the innate system. By his

suppression, limitation, and ordering, the child continually revises

all apsects of his system which distinguish his pronunciations from

the standards of adults around him -- until his system in fact becomes

the adult system. That massive variability may occur in children's

substitutions, or that the child may persist as long as he does in

apparently perceiving and reproducing his own unique system, are both

only apparent anomalies which offer no real contradiction of the theory

proposed. Each case can be readily enough explained: in the first

instance there is persistent correction of children's pronunciations


by surrounding adults, and in the second there is the lack of an

opportunity for the child to have conversation which would provide

the necessary feedback telling him whether or not his imitations have

been in any measure successful (Stampe 1973a).

1.2.4 The natural order of acquisition

Notable regularities can be observed (see Stampe 1973a; Salus and

Salus 1974; Ingram 1974) in the order in which children acquire and

master phonetic segments, and it is a proposition of Natural Phonology

that this natural order of acquisition is again readily explained by

the independently attested innate system. That is, the workings of

language acquisition will be fully accounted for by the hierarchical

arrangement and interactions of the natural processes.

A serious implication for child language theory as well as for

general phonological theory is the observation that Stampe's system

renders unnecessary Jakobson's hierarchical structural laws of ir-

reversible solidarity (cf. Jakobson 1968; Salus and Salus 1974; and

Stoel 1974). Stampe argues (in Stampe 1973c) that, like Chomsky and

Halle's "markedness" conventions, Jakobson's assumed implicational

laws which purportedly account for the system of language acquisition

(in children) and phenomena of language dissolution (in aphasics) are

mere appearances resulting from what is in actuality the underlying

system of innate natural substitution processes. Evidence for the

reality of active processes opposed to the possible reality of im-

plicational laws or conjectured reality of marking conventions hope-

fully will become more apparent as notions of Natural Phonology are

more ambitiously expanded throughout sections which make up Chapter

Three below.


1.2.5 The internal representations of adult speech

Although it is a fair assumption that the child passes through an

early stage of initial phonological development marked by incorrect

perceptions of an adult sound system (cf. Golick 1974), it is equally

apparent that mental representations which make up his own underlying -

system are very closely based on the adult models (Stampe 1969, Ingram

1974). Ingram (1974) expresses a view in line with Natural Phonology

when he suggests that "the mental representation (by the child) of the

adult model is not just the result of inadequate perception, but also

the result of organizational principles that the child uses to system-

atize this data" (pages 51-52).

An even stronger claim of Natural Phonology is that the child will

internalize representations of adult speech which are far more soph-

isticated than his own reduced reproductions would indicate. Stampe

has cited as one example his own son's rendering of "dog" as first

[da] (the same as for "doll" with no evidence of any final velar), then

[ga] (with the initial coronal assimilated to the deleted final velar),

and finally [gag] (with this final velar now pronounceable). Also

bearing on the issue is Jakobson's example, borrowed from Nadoleczny,

of the child who says in German first Duten Ta Herr Dotta, and then

Guken Gag Herr Goka, for "Guten Tag, Herr Doktor" (Stampe 1969, pages

446-47), another case which suggests strongly that children mentally

obtain phonological representations considerably before they can begin

to utter them.

Natural Phonology, then, has taken an extremely strong position

in the current controversy (cf. Kornfeld 1971) over whether the mature

adult system of phonological distinctions determines the child's system


or whether the child perceives as well as produces his own entirely un-

ique system of phonology. Stampe maintains the child's representations

are in terms of the adult's surface "phonemic" forms, which are inputs

to the system of natural substitution processes. This assumption is

a main subject of the discussion within Chapter Three which deals, among

other things, with Stampean phonemics versus classical phonemics. For

the present suffice it to say that the adult perceived-phonetic forms

(perceived, that is, by the child) plus the system of natural processes

(and eventually some "learned" rules as well) result in the child's

phonetic outputs. This particular interpretation can be and has been

expanded by Stampe into a Phonemic Hypothesis (cf. Section 3.1.2), which

claims that child and adult speakers alike deal most frequently with

shallow phonemic patterns much less abstract than the lexical represent-

ations envisioned to have psychological reality in Chomsly and Halle's

Sound Pattern of English (simply SPE throughout the remainder of the


1.2.6 The innate residue in adult grammars

The fully mature adult system of speech can be assumed, then, under

this framework, to retain all aspects of the innate system which this

mastering of pronunciation has left intact. Such a system maintains

as well the learned set of substitutions which are the phonological

rules and morphological rules also governing pronunciations -- though

not governing pronounceability. It is those innate processes which

fortuitously survive the course of language acquisition which actually

determine what are pronounceable (or at least what speakers assume are

the actually pronounceable) phonetic representations of the language.

For English speakers, the aspiration of initial voiceless stops is one


such surviving process, while the rampant deletion of final velars or

the devoicing of all final obstruents assuredly are not (though they

might be observed for a period of time in the pronunciations of many

of the children learning to speak English).

This last-mentioned example of obstruent devoicing is especially

suggestive and much has already been made of it in the literature on

Natural Phonology (esp. in Stampe 1968, 1969; Kiparsky 1968b). Thus

a highly-motivated process devoicing word-final obstruents appears

in many languages of the world almost simultaneously with the first

acquisitions of this type of segment. To master pronouncing the voiced

as well as the voiceless obstruents, then, the English-speaking child

must learn to suppress the process; but German-speaking children,

faced with a language in which such a process remains active, need

not. In fact, they must not suppress devoicing if they are to speak

acceptable German. In German this devoicing process turns out to

govern purely phonetic but not phonological environments, as the

always cited alternation of Bunt "association" with its plural Bunde

"associations" and dozens of similar forms reveal. With some few

languages (e.g. Hawaiian) it governs phonological representations as

well and there are no voicing opposition even on the most highly

abstract levels.

In this last case Stampe goes still further to contend that in

languages without any final stops (or even any final consonants at

all), such a process still remains part of the speaker's potential

inventory, a claim which he forcibly admits is unacceptable to any

other current system of phonology. But this unorthodox thesis

felicitously turns out to be enhanced by one small but very impressive


piece of evidence: speakers of these languages will regularly devoice

final obstruents once they are forced to confront these for the first

time in foreign loanwords, despite the total absence of any parallel

cases in their own dialects (Stampe 1969, page 445).

1.2.7 Naturalness motivations in linguistic change

In preceding paragraphs I have suggested something of the force

of Stampe's proposals in pointing towards more explanatory models for

synchronic grammars. More controversial proposals in light of the

traditional models are those which involve linguistic change, on the

one hand, and the phlnologically central role of borrowing phenomena,

on another.

A normative view of linguistic change is that rules may be added,

lost, generalized, or complicated in a grammar (cf. Miller 1973; King

1973, 1974; Dinnsen 1974). But one troublesome issue throughout has

been why rules should be added or complicated in the first place in a

process of linguistic change which is apparently motivated by the

pressures in languages toward overall simplification (Miller 1973).

In the theory of Natural Phonology advanced by Stampe, the system of

innate processes which accounts for language acquisition again serves

to explain as well these related events of phonetic change.

For Stampe, phonetic change results only when "the child fails

to suppress some innate process which does not apply in the standard

language" (1969, page 448). More generally, rule addition, generaliz-

ation, and unordering of processes are direct manifestations of the

child's failure to suppress (rule addition), to limit (rule general-

ization), or to order (rule unordering) natural processes with just

those three mechanisms mentioned for the revision of innate systems.


Such phonetic change is limited to the degree that the conservatism

of all speech communities assures the rejection of most innovations by

children. And when innovations in child speech are admitted in the

adult speech patterns, they usually are so only most gradually, i.e.

only as optional pronunciations.

There are additional reasons for opting for Stampean diachronic

linguistics. A view of phonetic change such as Stampe proposes has

the one special feature of accounting for exactly why change does not

seem to occur with adult speakers (except in some rare cases through

the influences of linguistic borrowing), why it is at times quite

radical, but why at the same time it is always astoundingly regular.

For, "if a child fails to master a certain sound, it will appear that

he has changed it to the sound he regularly substituted for it"

(Stampe 1969, page 448). Notice that this view of historical change

refers only to natural phonological processes and makes no references

whatsoever to learned and unproductive phonological "rules". This

is a feature of another current version of natural phonology as well

-- Vennemann's Natural Generative Phonology. All new rules are

phonetically motivated and clearly have their source in some universal

inventory of phonetic constraints or conditions. Rules are simply

not added to grammars when they have grammatical conditioning or when

they reveal surface exceptions. Both Stampe and Vennemann draw a

similar (though not identical) distinction between the natural phono-

logical processes (the "phonetic rules" for Vennemann) and acquired

phonological rules (the "morpho-syntactic rules" for Vennemann) within

the amorphous class of substitution processes lumped together in a

standard generative theory as the "phonological rules" (see Chapter


Eight for fuller discussion of Vennemann versus Stampe).

Stampe's view that phonetic change follows only from the child's

failure to suppress some natural process which should not apply in the

standard language, and therefore that it does not involve learned rules

at all, is to be sharply contrasted with a view of Jakobson, Halle,

Kiparsky, and others, that change is also motivated by addition and at

times suppression of more cognitive phonological processes. I explore

this distinction further in Sections 3.1.1 and 3.1.3 within Chapter

Three and give there three particularly illustrative examples based

on current literature (Daniels 1973; Cressey 1974; Dinnsen 1974) of

how phonetic change might be more insightfully interpreted in the

framework of a Natural Phonology than in the framework of a standard

generative phonology. The position of Stampe versus that of Halle or

Kiparsky should, of course, be open to an empirical verification and

such verifications will also be taken up in Chapter Two and Chapter

Three (as will the empirical support for Stampe's model over that of

Vennemann). In essence, Stampe contends that the few recorded genuine

cases of the suppression of natural processes (e.g. loss of final

obstruent devoicing from Yiddish and Swiss dialects of German) are

exceedingly rare and can at any rate be sufficiently explained away

by a complex of factors that seem to have nothing at all to do with

the pressures toward grammatical simplification (see note twelve for

Chapter Three). The result is clearly that suppression may be listed

among the possible but not among the primary mechanisms of language


1.2.8 The prominence of natural loanword phonology

Stampe 1969 asserts byt makes no effort to demonstrate that "the


problem of 'phonological admissibility' is contained in, and therefore

inseparable from, the larger problem of loan phonology" (page 451, note

4). To a degree at least this present study will attempt just such a

demonstration, offering one of the more extensive although certainly

not the first investigation along these lines.

Earlier though woefully incomplete attempts to establish the

intimate connections between principles of Natural Phonology and the

events of loanword phonology are represented by the work of Mieko Ohso

(1971) and Julie Lovins (1973, 1974a, 1974b). Especially in Lovins

1974b conclusions are drawn from Japanese lexical study which would

seem to support the contention that Stampe's view of an innate system

of processes explains events of linguistic borrowing as readily as it

accounts satisfactorily for events of infant speech or processes of

linguistic change. Since they motivate as well as sustain my own

observations about borrowings, Lovins's conclusions are valuable to

quote here in some detail, precisely as she has arranged them throughout

Lovins 1974b. These proceed as follows:

(1) Perception of the "closest sound" in another language occurs
in terms of phonological processes, not binary features: features
are not perceptual primes (page 242).

(2) Systematic non-uniform criteriality of features, dependent
on the phonetic substance of segments, corresponds to this large
inventory of processes. This also explains the existence (though
not the distribution) of alternative substitutions, since more
than one process may apply to a given segment (page 242).

(3) Processes governing the phonological system of a language
sometimes appear explicitly only in an interference situation.
Often those processes constraining underlying representation will
determine the "closest sound" (page 242).

(4) Allophonic distribution must be considered in characterizing
the "closest sound"......Sounds are often perceived sequentially
not individually, and in relation to context-sensitive processes
(page 243).

(5) A context-free process determines perception of a foreign
sound only if no relevant context-sensitive process is available to
do so7 (page 243).

(6) Loanwords have underlying representation at the phonemic level
.....Since morphophonemic processes determine underlying represent-
ations only when this is required by specific morpheme alternations,
and loanwords are not involved in such alternations, they are auto-
matically represented phonemically. This fact has sometimes been
remarked on as one aspect of the exceptionalityy" of loanwords, but
this type of "exception" is merely a demonstration of the over-
abstract underlying representations being forced upon non-foreign
morphemes in Language T by linguists (where "T" is "target" language
-- PCB)8 (page 243).

(7) When we speak, we apply allophonic processes "forwards" to
produce contextual variants; when we listen to someone else, we apply
.them "backwards" to relate the allophones to their associated phonemes,
automatically considering a sequence of as many segments at a time
as are involved (as context or derivates) in a derivation. Likewise,
in listening to unfamiliar foreign sounds, we try to relate what we
hear to possible surface forms in our language. These surface -
forms may already be acceptable underlying representations, or re-
lated to such by backwards-derivation of an allophonic process......
In the "neither" case, already mentioned, one has two options: apply
a context-free process governing underlying representations, to
get an acceptable one; or backwards-derive the segment up to the
point where it is a phonological exception before doing this (i.e.
before applying the context-free process -- PCB) (page 244).

(8) The "environment" part of a context-sensitive process may be
generalized in the perception of a foreign sound sequence......
Stampe has pointed out that this is what happens when we listen to
"baby talk": a yound child's phonology is different from ours large-
ly in that a lot more processes apply, or they apply more generally;
in order to understand what the child is saying, we have to relate
what is normally the output of our derivations to his less restricted
phonetic forms. To the extent that we can still do this we can
correctly interpret what he is saying (page 244).

(9) When a foreign segment appears in an environment in which the
equivalent native derived segment does not appear, then the form
of the incoming foreign word is modified so that the structural
description of that rule is met and the segment in question is then
derived in the appropriate environment (page 246).

(10) When a sequence of foreign sounds not corresponding to a
legal derived sequence in the target language is interpreted
according to a context-sensitive process, the environment of this
process will be generalized to the smallest natural class that
encompasses the corresponding segment in the source sequence (n.b.
this is a less formal restatement of the previous theorem -- PCB)
(page 246).


(11) The "degree of nativization" of a loanword is determined by
what processes have applied in its lexicalization, if there is a
choice; and what rules, if any, have applied after its initial
lexicalization (leading to relexicalization) (page 248).

(12) "Native" forms are those to which all (or most?) rules have
applied, as applicable. "Assimilated" loans are those to which all
processes have applied. "Non-assimilated" loans contain either a
violation of a process constraining underlying representation (are
phonological exceptions) or of one that normally determines surface
representation (are phonetic exceptions) (page 248).

As it is codified and summarized in these dozen theorems, then,

Lovins's work takes an ambitious initial step toward elucidating some

conditions under which each of two important processes responsible for

lexicalization in Natural Phonology apply in the course of lexical

borrowing. These are the forwards application of the higher-level

morphophonemic processes and the backwards application of lower-level

allophonic ones. Tentatively at least, Lovins provides the first

formulation of a significant principle: that "at almost all points the

elusive concept of 'closest segment' in two languages is clarified in

terms of these two conditions" (1973, page 2). A more general but

nonetheless equally significant outgrowth of Lovins's study of foreign

borrowings would also have to be the observation that innate pro-

cesses as Stampe has conceived of them maintain the most vital bear-

ing on loanword phonology, with the role of "rules" in borrowing still

being largely indeterminant though most likely not overly productive.

It seems reasonable, then, that Lovins should assume that looking

for an adequate theory to account for the behavior of loanwords in

essence is equivalent to "arguing for the efficacy of a theory of

natural phonology such as proposed by Stampe......and arguing against

some tenets of nonrevisionist generative phonology" (1974b, page 240).

However, a weakness of Lovins's work (as for that matter of Stampe's)


is the absence of any widespread empirical verifications, being based

as this research is on the necessarily limiting data from a single source

language. Natural Loanword Phonology (Natural Phonology in general)

remains, for most linguists familiar with it at all, an untested theory

which raises considerable scepticisms and even numerous hostilities.

Part of the reason no doubt has been advanced by two recent commentaries

which suggest that Stampe's collective theory is not only untested but

perhaps virtually untestable as well (see Stoel 1974 and Ohala 1974).

This has from the first been an altogether familiar and unfortunate

categorization of Stampe (n.b., however, Ohala's retraction quoted

below in note 9). It is as rebuttal to particularly this estimation

of Stampe (the untestability hypothesis) that I have pressed my own

conviction here: i.e. that it is not only possible but even eminently

reasonable to assume that such hypotheses about loanword phonology as

those formulated by Lovins and those derivable directly from Stampe

(1973a) are not only susceptable to testing but even highly comparable

with empirical verification or disconfirmation -- and along with them

the general tenets of Natural Phonology as well. While much of this

dissertation (Part One and Part Two) is of necessity devoted to merely

explaining Natural Phonology, in concluding chapters I have exposed

Lovins's hypotheses, along with others which derive directly from

them, to extensive additional analysis in light of renderings of

numerous English loanwords in the Miami Cuban dialect of Spanish.

One welcome result of Chapter Seven and Chapter Eight, for example,

is the expected conclusion that loanword phonology is not only con-

siderably clarified by application of Stampe's theories but also it-

self supplies in return what seem to be the necessary verifications


for Natural Phonology, proofs which are otherwise not always apparent

from related areas of exploration in natural languages -- such as child

phonology, speech errors, or observable evidence from the adult system

of phonology and especially from casual speech phenomena.

To review briefly, what I have given in this section is a summary

of the foundational claims about Natural Phonology. I have been most

concerned here with emphasizing (1) the innate system of phonological

processes; (2) the physiological and mentalistic motivation of these

processes versus the abstract and highly unproductive status of distinct

"rules"; and (3) some implications, at least, of Stampe's notion of

processes for possible interpretations of phonetic change and of a

synchronic loanword phonology. But what is presented here is at best

a superficial outline of only some of these hypotheses fundamental

to Stampe's theory. Throughout the first half of this present work

these claims are more thoroughly explored and, to whatever extent

possible, empirically motivated. Throughout these earlier chapters,

I will attempt to view Natural Phonology primarily from the standpoint

of the reality and productivity of the innate processes, both as they

operate in historical change through time and as they function in the

acquisition of individual grammars. Eventually, of course, the vested

interest of this dissertation will be in the specific issue of loanword

phonology and in particular cases of contact borrowing within a single

synchronic grammar, and these will be taken up in detail with Chapter

Five, Chapter Seven, and Chapter Eight.

1.3 Explanation and Performance Models

At the forefront of Stampe's work on Natural Phonology is the


assumption that our recognition of an innate system of processes will

make available to us explanations for linguistic behavior not possible

with the standard generative phonological theories. At least not as

these theories, which might be referred to collectively as "abstract

generative phonology" (Terrell 1975d), have been previously conceived.

The subject matter of the few existing monographs and articles on

Natural Phonology produced to date has persistently involved some

attempts at just these types of explanation, like those we have already

hinted at for language change and linguistic borrowing, for example.

Another topic prevalent in the writings on Natural Phonology

(esp. in Stampe 1969, 1973a) has been the considerable evidence that

innate physiological and mental processes are, in point of fact, true

substitutions actively applying in psychologically real grammars. Yet

to date the bulk of these discussions focus on evidence from child

language research which, as we have just noted, is only one (though

perhaps the most exploitable one) among several obvious testing

grounds. The work which we have earlier alluded to (Section 1.2.8)

by Lovins seems sufficient to suggest that the observable loanword

material is equally well adaptable to this framework and therefore

equally supportive of Stampe's principles. But we have just finished

claiming that Lovins's studies in another sense only serve to emphasize

for us the huge gaps surrounding existing studies on loanword phonology,

and it is with the hope of filling some of these gaps and defining

some of the intractable issues that my own investigations of phonetic

borrowing in Cuban Spanish have been undertaken.

It is to be borne in mind then, throughout this and other writings

devoted to Natural Phonology, that Stampe's version of "naturalness"


takes the goal of a fully adequate linguistic theory to be one of

"explanation" in phonology. This is a marked departure from the early

goal of Chomsky, which is unambiguously stated to be no more than "the

construction of a grammar" (Chomsky and Halle 1968, page 3), where

a grammar of a language "purports to be a description of the ideal

speaker-hearer's intrinsic competence" (Chomsky 1965, page 4).

To establish Stampe's contrasting commitment to "explanation" as

the true and single goal of phonology, it is necessary only to point

to some recent assessments he has made of Vennemann. One unpublished

observation, for example, rejects quite out of hand Vennemann's

label of "natural generative phonology" because it misses precisely

the point of a distinction between Natural Phonology and generative

phonology: that "even at the level of universal grammar generative

phonology is a descriptive rather than an explanatory enterprise"

(Stampe personal communication). Stampe takes "generate," then, i.e.

in the sense employed by Chomsky, to mean more narrowly "to describe

all and only" (Stampe personal communication). The broader goal of

any linguistic theory for the natural phonologist is rather "to

understand language, to see how its peculiarities follow necessarily

from the nature of its function, its implementation, and its use"

(Stampe personal communication).

In light of Stampe's own careful dichotomy between "natural" and

"generative" as distinct and even irreconcilable approaches, it would

seem essential that we emphasize foremost here Natural Phonology (as

so far proposed by Stampe, by Rhodes and Patricia Donegan Miller, and

in my own papers in Bjarkman 1974a, 1975) as a radical departure from

standard generative linguistics. In several important senses Stampe's


model is clearly a completely different theory from Chomsky's, rather

than perhaps a close notational variant of those types of phonological

analysis envisioned in The Sound Pattern of English or adopted by Postal,

McCawley, Kiparsky, or most other orthodox transformationalists. (One

overview of this uniqueness of Natural Phonology is now to be found

in Bailey 1975.) For one thing, I will argue extensively in Chapter

Three that Stampe's concept of natural processes is not at all reducible

to any SPE notion of the late phonetic rules. This is largely of course

because Natural Phonology, as already underscored, pays little attention

to any categories among the "acquired" rules, other than in the negative

sense of assuming that most rules proposed in SPE for English have the

status of processes and are not "rules" at all in the Chomskyan sense of
the term. That is, the division between rules and processes occurs

(or at least is envisioned by the linguist) deeper in the grammar than

a systematic phonetic level; or put otherwise, processes are not simply

detail rules with n-ary feature scales but have in many cases true

morphophonemic as well as allophonic significance for the grammar.

Yet the real departure of Stampe's version of Natural Phonology is

that within a general framework which can still be considered loosely

to be generative grammar it makes giant strides toward achieving a true

performance model of phonology.

Generative grammarians collectively have remained largely uncertain

about the role of "performance" in the idealized formal grammars they

write. Therefore one recent study, by Derwing (1973), is for just this

reason a considerable achievement, if only in the sense that it brings

about some lucidity in Chomsky's otherwise confusing and shifting stance

on what would constitute a competence/performance distinction in the


interpretation of generative grammars. Derwing unravels Chomsky's

final position to be essentially uncompromising in his opposition to

applying strict performance criteria:

Transformational-generative grammars are not intended to be
interpreted as idealized production and/or perception models, after
all [e.g. after the fashion of Saussure's famous metaphor that a
language is comparable to a symphony and speech to the performance
of that symphony -- PCB], but as models of something else "neutral"
between the two; in other words, a model of "linguistic competence"
(as this "something else" has since come to be known) is not to
be regarded as an idealized model of linguistic performance
(Derwing 1973, page 265).

In other terms, while Chomsky repeatedly speaks of the desirability of

models to test linguistic performance, he strenuously avoids the close

interpretation of generative grammar in this vein. And Chomsky himself

again echoes this point with his restatement in Aspects of the Theory

of Syntax:

To avoid what has been a continuing misunderstanding, it is
perhaps worth while to reiterate that a generative grammar is not
a model for a speaker or a hearer. It attempts to characterize
in the most neutral terms possible the knowledge of the language
that provides a basis for actual use of language by a speaker-
hearer......When we say that a sentence has a certain derivation
with respect to a particular generative grammar, we say nothing
about how the speaker or hearer might proceed in some practical
or efficient way to construct such a derivation. These questions
belong to the theory of language use -- the theory of performance
(1965, page 9).

In light of such passages it is all the more significant that Stampe

has begun his own discussions of the organization of a system of natural

processes with the observation that "there has been a tendency in lin-

guistics to view processes of grammar as descriptions of the language

'competence' of speakers, and not the actual processes that occur in the

production or perception ('performance') of speech" (1973a, page 43).

Thus inevitably, "explanations of linguistic phenomena based on 'as-if'

descriptions remain 'as-if' explanations" and, what is perhaps worse,


miss the vital generalization that "the conditions of the use of language

(performance) are responsible for the nature of language" (1973a, page

43, with Stampe's own underlying). As again Derwing (1973) emphasizes

with a special urgency, it is this absence of a performance grammar

which is the peculiarly debilitating feature of the SPE notion of the

phonology.11 McCawley notes of Derwing (and thus indirectly about the

transformationalists that "he is particularly critical of the failure

of transformational grammarians to concern themselves with performance

to any greater extent than merely formulating programs for its inves-

tigation" (McCawley 1974, page 178). Natural Phonology is by stark

contrast a genuine performance model, if only to the degree that it

takes the psychological reality of substitution processes rather than

economy criteria and simplicity measures as the proper constraint on

highly-valued grammars. The natural phonologist does not exactly

advocate that linguists set out to write grammars which are uneconomical,

though he would presumably advocate that they more consistently use the

data from linguistic change and from psychological experiments to try

to uncover exactly what grammars are really like (Darden 1971, page


A second acknowledged and not altogether unrelated weakness of

generative phonology as it stands has been the absence in an SPE model

of any true theory of lexical representation. Once again it is Derwing

(1973, pages 127-28) especially who has captured the subtle implications

of such oversights:

A related, and equally disturbing, development in current
phonological theory is that underlying representations for mor-
phemes appear to be getting more and more like the conventional
orthographic ones......But if this is so, we are faced with a
strange anomaly: why do the problems encountered by English-


speaking children in learning their own "near optimal" spelling
system (in which the orthographic representation of morphemes cor-
responds very closely to their abstract lexical representations in
a generative grammar) far outweigh the minimal difficulties which
the typical Russian child appears to have in learning to spell
Russian (where the orthography diverges greatly from Lightner's
level of lexical representation, and parallels instead the less
abstract morphophonemic representation of the sort proposed by

The causes are of course obtrusive at this point. Yet part of the

reason at least would seem to follow from the cogent observation of

another recent critic of nonrevisionist generative grammar, Mary Louise

Clayton, who contends that a major flaw in the theory is that "there

is no clear principle establishing the degree to which a phonological

representation can legitimately differ from its surface representations)

on the one hand and its lexical representation on the other" (Clayton

1974, page 155).

Clayton would seem to relate this anomaly to the fact that "an

assumption apparently arose generally in the field that only phonolog-

ical rules were necessary to relate surface representations to under-

lying (lexical) representations" (1974, page 160). The flaw in such an

assumption is clearly that it "confused the issues of the abstractness

of phonological representations, which is a question of the nature of

phonological rules, and the abstractness of lexical representations,

which is a question of what alternations should be accounted for by the

grammar" (page 160).

Clayton concludes her assessment of the issue on a note of surprise

that, although Kiparsky (1968a) takes up the "abstractness" question,

a notion of "strictly phonological" as a desirable restriction on what

should be admitted as "phonological rules" (the topic of her own study)

has not surfaced until works of a much more recent vintage. But of


course none of this would be particularly surprising is we recall that

until very recent and unprecedented work by, for example, Halle (1973)

and Skousen (1973, 1974) there has been no approach which allows for

"morphological" rules which are in any substantive way distinct from

the amorphous phonological component which considers all "phonological"

substitutions to be quite identical and inseparable by nature (compare

e.g. Schane 1971).

The failure to provide a convincing theory of lexical representation

within generative grammar should in itself serve as a sufficient motive

for the goals and organization of this present study. VWereas'in the

first third of this thesis I undertake describing Natural Phonology in

terms of a performance model of phonological grammar, with Chapter Four

I also make some tentative first proposals about what a lexicon must

look like for an adequate natural generative grammar. These are pro-

posals which combine the standard notions of grammatical components

(syntactic, morphological, lexical, phonological, and semantic) and

lexical insertion with Stampe's views of what comprises a natural sys-

tem of phonological rules. My model is discovered to be much like that

found in Halle 1973 or implicit in Harris 1974. It assumes that there

is for each speaker a dictionary of lexical items which results from

applications of word formation rules and an exception filter on lists

of formatives (morphemes) and that it is within this separate notion

of Morphological Component (i.e. before the fact of lexical insertion)

that many of the alternations previously taken for "phonological" rules

will actually apply.

A superficially similar proposal for investing generative grammars

with independent morphological rules is already to be found in the


miscellaneous work of Royal Skousen (1972, 1973, 1974). For Skousen,

however, a phonology is made up of rules which account for two distinct

kinds of output regularities: (a) fully productive rules which are

phonetically motivated and always entirely exceptionless in their ap-

plications; and (b) rules of "morphological alternations" which are re-

placements in the synchronic grammar for rules that at one time were

(but now no longer are) phonetic and exceptionless. (Skousen's rather

impressive evidence that speakers do not always capture underlying regu-

larities that are products of the historical rules turns out to be a case

where speakers of Finnish change new words to respond to surface patterns

of Finnish by corresponding with other forms undergoing the historical

rules, even though this means directly contradicting some phonetically-

plausible rules.) One immediate problem for Skousen's analysis (when

opposed to that of standard theory or that of Halle's ideas of word

formation) is that it makes no apparent provision for the vast majority

of rules which are neither phonetically-plausible nor attested by mor-

phological alternation and which comprise the largest class of rules

generally called phonological rules by SPE. In Skousen 1973 this

difficulty is seemingly resolved in a most curious way. The solution

is to have morphophonemic alternations (those relating what generative

phonologists call the systematic phonemic and systematic phonetic repre-

sentations) accounted for by morphological rules which may well be

phonetically-conditioned but do not stand as phonetically-plausible

generalizations about the grammar. This third category of rule comes

about when a number of morphological rules state over and again an

identical alternation in the identical phonological environments.

Skousen proposes that speakers, then, are able to recapture historical


environments for "dead" rules like e.g. Gemination in Finnish precisely

because "the process of postulating morphological rules leads directly

to the original rules......the evidence for the phonetically statable

environment is discovered by an acquisition procedure that depends upon

morphologically-defined surface rules" (1973, page 25). It remains

undetermined with Skousen's theory what kind of a "level" is relevant

for stating morphological rules, though it is suggested that "a theory

using morphological rules implies, in fact, that underlying represent-

ations are not very deep at all" (page 22) and also that analogicall

change suggests that the linguistically significant level in phonology

is the surface phonemic level" (page 22).

In Skousen's papers we uncover a phonological system that sounds

vaguely akin to that outlined for Stampean phonology (see Section 1.2)

earlier within this chapter: a traditional phonological component is

reassessed as comprising learned non-productive generalizations (for

Skousen morphological alternations) and exceptionless phonetically

plausible rules, and a linguistically relevant surface phonemic level.

Skousen's morphological rules are, from what he reveals about them,

what taxonomic phonemics would call morphophonemic processes, while

his phonetic rules approximate the allophonic processes of the

structuralists. Such a format allows distinctions (actually it

claims that speakers make such distinctions) not available in an

SPE notion of phonology, where speakers are assumed to have one

underlying phonological level (the systematic phonemic or lexical

level) and where it is also assumed that a single type of rule

will derive from this level both the morphophonemic and allophonic

alternations without any provision for distinguishing which are of

which type.

But like other recent arguments for restricting phonological rules

to productive phonetic rules (see Clayton 1974), Skousen's model is

only vaguely defined and therefore not readily testable. Most of what

follows is devoted to arguing advantages of a modified version of the

Stampean proposals over a partial and largely unempirical system like

Skousen's. What is most problematical about the latter is that if

Skousen recognizes distinct types among phonological alternations,

he never indicates how actual speakers might ever do this, nor how

deep the "not very deep" underlying representations are and by what

principles this could be determined. An insurmountable difficulty for

Skousen (though not at all one for Stampe) seems to be also that the

morphophonemic rules are sometimes actually phonetic in motivation

(see my discussion of German final devoicing processes in Section 2.3).

The notion of morphology suggested here in Chapter Four, on the other

hand, is one in which the morphological rules are word formation

conditions within a distinct but "passive" morphological component

(i.e. the lexicon) and thus remain outside altogether the province

of active derivations containing morphophonemic or allophonic pro-

cesses. Morphological rules constitute a truly "passive" grammar --

in the sense that the speaker only applies them when needed to account

for some new or unfamiliar form. It is this idea of morphology as a

separate component of the grammar which makes possible explanations

for rather intractable problems in earlier approaches to loanword

phonology; as a case in point, it is surprisingly easy to demonstrate

that Cearly (1974) is irrecoverably wrong when he asserts that phono-

logical rules apply indiscriminately to foreign forms but (what he


calls) morphological rules never manifest themselves in this way (see

Section 7.3 below). The idea of a "passive" grammar also proves to

be advantageous in accounting for how such rules can seemingly be

inserted at random points througholiL a derivation, in apparent con-

tradiction of the generally accepted ordering principle of first the

morphological, then phonological, and finally phonetic rules. The

sustaining of such an ordering principle is a difficulty earlier

noticed (especially in Skousen 1972) and both Skousen and Anderson

(1974, 1975) provide ample illustration of phonetically conditioned

rules followed by morphological rules (e.g. in English the rule

governing a/an distribution applies to outputs of initial h deletion),

a case which proves embarrassing to any notion of linearly-ordered

strings of morpholexical, phonological, and phonetic rules. My own

conception of a phonological component (Section 4.2.0) would seem to

successfully avoid such perplexities in rule ordering.

I have given some attention to Skousen's approach to morpho-

phonology here since it appears to represent a single serious attempt,

outside of those of Halle and Harris, to formalize a morphological

component within a generative grammar. Further studies in this vein

are exceedingly rare and often only obliquely related to the topic

at hand.

Saciuk's dissertation (Saciuk 1969), for instance, provides an

exceptionally ambitious effort to study "the role that the components

of the lexicon play in generative phonology, and to sketch a proposal

for dealing with these phenomena" (1969, page iv). Like several known

predecessors (and he cites precursors ranging from the Prague School

of Mathesius through the M.I.T. School of McCawley and Harris) who


have remarked about dividing a vocabulary into native/non-native

suncomponents, Saciuk takes loanwords and the other "exceptional"

elements as given facts of a language (in this case Spanish, Portuguese

and Catalan) as it is attested in the single speaker; he concentrates

not on the workings of "nativization," but on the role of different

hypothesized native and foreign components of "lexical strata" within

a lexicon which has become a fixed synchronic feature. Saciuk 1969

serves, then, like its earlier prototypes, as a proposal for describing

the phonology of a target language after the established fact of lan-

guage contact. However it has nothing to say about what are the actual

processes of lexicalization.1

Chapter Four of this study in particular departs from generative

studies like Saciuk 1969 in taking as its focus the role of the lexicon

(i.e. rules of word formation which in a model approximating Halle's

give structure to formatives comprising the lexical listings or stored

dictionary) in actually restructuring new loanwords in early stages

of borrowing.

Some limited attention is also given in Chapter Seven to methods

for eventual nativization of foreign segments as they become over time

fully a part of the native system. The intimate connections between

Natural Phonology, loanword phonology, and the structure of a speaker's

lexicon should of course be patently obvious: the structure of the

lexicon is product of a system of phonology which itself is revealed

transparently by (among other various types of alternations) the manner

in which borrowings are handled by the native speaker. After all, when

we inquire how processes are active in determine our perception of

foreign sounds it is equivalent to inquiring how they determine the


perception of our own native system (see Lovins 1974b). And any com-

plete loanword phonology must deal not only with initial treatment of

borrowings in the native system but with questions as well of how

foreign elements eventually lose their exceptionalityy" and become

fully and irreversibly "nativized." It is not difficult to conclude

that a study of the processes of borrowing reduces inevitably to a

study of the processes of lexicalization. And such conclusions lend

weighty support to the theory Stampe espouses: that all elements of

phonological systems -- diachronic and synchronic, developmental (child

language) and degenerative apraxiaa), native and foreign -- derive

from the single underlying principle of innate natural processes.

Eventually, of course, full and scientific treatment of the lexicon

(word formation processes and not just allomorphy) from the standpoint

of Natural Phonology will be altogether indispensable, and it is a

first step in precisely this direction that I have aimed for with

Chapter Four.14

Though little has been said to date within Natural Phonology about

the subject of lexical representations, certain implications of this

theory for the structure of a lexicon are also somewhat inevitable.

One immediately obvious distinction between the concept of the lexicon

required by Natural Phonology and that in standard generative theory

might help in sorting out contrasting tenets of the two theories. This

is that whereas in abstract generative phonology the lexicon is over

and again employed as a type of repository for all problems resulting

from an over rigid simplicity criterion (e.g. an inventory of segments

is proposed on the basis of Markedness criteria and features counted,

along with principles of symmetry, and the more abstract these segments


prove to be the more complex and numerous become the rules needed

to account for them), in a system of Natural Phonology (which assumes

processes to be the primitives) the content and organization of the

inventory of segments is never analyzable in isolation but instead is

always a manifestation of the process system itself (compare Lovins

1973, page 27). Since both immediate lexicalization and eventual

nativization result from either forwards application of higher-level

processes or backwards application of the lower-level ones, knowledge

of the processes of the native system provides a method for both

speaker and linguist to arrive at proper underlying forms (a conven-

ience not available to either Skousen or Chomsky and Halle). Implica-

tional laws of Jakobson, Markedness of SPE, redundancy conditions-of

Stanley, MS conditions of Halle, and surface constraints of Shibatani

-- all these reduce to "an innate system of phonological processes

which resemble the implicational laws and the markedness conventions

in content but have the same ontological status as the native processes

(so-called "rules") of the phonological system of any individual lan-

guage" (Stampe 1973c, pages 44-45). It is in this particular sense,

then, that conditions on lexical representation must be dealt with in

subsequent chapters. And it is also in this special sense Stampe's

notions of processes may be shown to be entirely as relevant in

accounting for the nature of a speaker's lexicon as they are in ex-

plaining observable surface phenomena of acquisitional phonology, or

as they are in describing the seemingly related substitutions which

predominate with examples of casual or unguarded speech and even with

examples of drunken speech (Lester and Skousen 1974).


1.4 Remarks on Cuban Spanish and Methodology

No assumption is made with subsequent chapters that readers of

this dissertation are previously familiar with any precepts of Natural

Phonology beyond those few now sketched in Section 1.2. Quite to

the contrary, this general unfamiliarity makes Part Two of necessity

a lengthy account of Stampe's theory: its workings and its basic


In order for these initial chapters to be intelligible, however,

and in order that familiar positions and controversies of generative

phonology (e.g. Markedness, abstractness, extrinsic rule order, and

motivations for linguistic change, etc.) need not be entirely restated

fresh, it must be assumed that any reader entering here is thoroughly

versed in these fundamental "issues" of generative theory. That is,

a minimum acquaintance is assumed with the important books as well as

articles and circulating papers of at least the following generative

phonologists: Chomsky and Halle (especially SPE), Postal, McCawley,

and Kiparsky (e.g. his 1965 dissertation Phonological Change and

milestone 1968 article "How Abstract is Phonology?").

At least a basic working familiarity with generative treatments

of Spanish phonology is also taken as granted. Observations to be

made generally about the phonology of the American Spanish dialects

are based on the seminal book by Harris (1969) and the dissertations

of Foley (1965) and Saciuk (1969) -- all three featuring the Mexico

City dialect of Spanish as a standard -- and it is expected that any

readers have also absorbed these earlier treatments as background.

Since previous works on Cuban Spanish are rare, on the other hand, all

claims about these dialects are more fully documented from original


and sometimes rather obscure sources.

Finally, assumptions about phonology in Miami Cuban Spanish (MCS)

are drawn from my own field work, as well as from studies which are

now in progress or which have very recently appeared. One such work

is an ongoing analysis of MCS phonology by Bohdan Saciuk, which has

thus far resulted in only one formal presentation (Saciuk 1974) yet

promises significant contributions to American Spanish dialectology.

The second is a recent dissertation by Robert Ilammond (1976) which

undertakes the study of certain rapid speech processes of MCS and

broaches a discussion of implications of such processes for universal

phonological theory. Hammond's dissertation, especially, suggests

some crucial topics for debate when we turn to a specific analysis

of MCS in Chapter Five and Chapter Seven of this present study.

Data gathered about the Cuban dialects until recent years

remained remarkably sparse. And where it was available it was

.generally less than properly analyzed. Often there are conflicting

observations and even contradictory conclusions.

As examples, studies by Isbasescu (1965, 1968), based one a

handful of informants who were Cuban students resident in Bucarest,

suggest that a velar nasal allophone appears generally before k

and in utterance-final position, though occasionally before other

consonants as well. However, in Guitart 1973, which is drawn from

informal observations of fellow Cuban exiles in the United States

and supposedly represents an Educated Spanish of Havana (ESH), we

discover that at least for Guitart's personal dialect (though I am

not convinced this is adequate generally even for his own personal

idiolect) the obligatory rule is for all closed syllable-final


nasals to be velars. Though Saciuk (1974) notes a strong tendency

among Habaneros to voice intervocalic stops, this tendency has received

only scattered and fleeting reference in the earlier studies. More

recently, Hammond (1976) labels as "rapid speech phonological processes"

velar nasal intrusion, intervocalic voicing of the voiceless stops,

s-aspiration and deletion, r/l confusion, devoicing of final unstressed

vowels, final consonant deletions, spirantization of voiceless stops,

monophthongization, and others. Yet monophthongization is a morpho-

logical rule rather than a phonological rule and its relationship to

rate of speech is at best highly questionable. And concerning most if

not all the remaining processes mentioned, while they are undoubtedly

more frequent in more casual speech and thus in some sense tied to

speech style, still any causal relationship to speech rate would be

prohibitively difficult to verify. My own observation is that all

these processes are frequent though random in occurrence even in the

most guarded style of speech reported.

Chapter Five will treat these issues generally, and in particular

the problems of spirantization and what Guitart (1973) and Hammond

(1976) both record as velarization of nasals will be scrutinized. To

merely hint for the present at what seems an essential inadequacy of

Hammond's concentration within a generative phonological framework on

"rapid speech phonological processes" cited above, we might observe

that spirantization and final consonant deletions, at the very minimum,

represent certainly optional and sometimes obligatory rules in the

non-casual as well as casual speech for most speakers. To suggest

that these "rules" also extend their effects to sloppy or careless

speech patterns is only to suggest that they are in reality universal


phonetic processes (in the Stampean sense) and not acquired "rules" at

all. Also, that such processes spread their effects to casual speech

is only endemic to their character as natural processes (see Section

5.4 below) and not any indication that they might be at all equivalent

to an SPE type notion of the phonetic detail rules.

An earlier scarcity of work on Cuban Spanish is recently being

replaced with a surprising burst of scholarly interest and energy. In

addition to Guitart 1973 and Hammond 1975, 1976 there are two further

dissertations of the Spanish of Havana (Lamb 1968 and Sosa 1974) and

a limited but steadily expanding list of articles on Cuban phonology.

Guitart 1973 remains, however, the only monograph-length contribution

on insular Cuban Spanish which merits careful attention with regard to

my own study.15 Of considerable significance as well (though more as

a spur to further debate than as any empirical solutions to reigning

problems in Cuban phonology) are a series of unpublished and soon-to-be

published papers by Terrell (see Terrell 1974a, 1974b, 1975a, 1975b,

1975c, 1975d). These papers, with a theoretical orientation toward

Vennemann's Natural Generative Grammar, treat the following topics:

(1) functional grammatical constraints on deletion of word-final s

in Cuban Spanish; (2) percentages of occurrence and the interrelation

of processes of aspiration and deletion of syllable-final and word-final

s in Cuban Spanish; (3) percentages of occurrence and interpretation

of implosive and final nasal phones; and (4) evidence from the above

phenomena in the Spanish of Cuba which lends support to an interpre-

tation by means of Vennemann's natural generative theories. Data

utilized by Terrell is from a series of taped interviews of twenty-two

Havana natives carried out in Miami as part of the "Coordinated Study


of the Linguistic Norms of the Principal Cities of Ibero-America and

the Iberian Peninsula" (Comision de Linguistica Geografica, Consejo

Superior de Investigaciones Cientificas, Madrid).16 Unlike my own in-

formants, however, none of those referred to hy Terrell had been in the

United States for more than three months at the time of the interviews.

I will not review the literature on Cuban phonology in any detail

in this current chapter, as this has been the practice of most previous

dissertations and especially thorough reviews are already available in

chapters by both Lamb and Hammond (though Lamb's review unfortunately

antedates Terrell's work and a number of other important papers). We

might only observe that studies on Cuban Spanish have fallen largely

into three distinct categories.

First, there is the major category of studies which consists ex-

clusively of lexical compilations, mostly produced early in the present

century. These are of little use to contemporary phonologists yet they

have been unfortunately almost a sole basis for a number of earlier

phonological studies (or reputed phonological studies). Lamb 1968 gives

a brief account of such compilations in its second chapter reviewing

the history of Cuban phonological studies.

Second, there are two works of major proportions done from a

structuralist standpoint and describing the Spanish of the Havana

Province. Lamb 1968 is itself a traditionalist description of phonemic

and phonetic inventories with some additional discussion of general

characteristics of the Caribbean linguistic zone; Sosa 1974 attempts

to relate Cuban phonology to a Spanish Creole and to trace features of

a Cuban dialect to the influence of slave populations over the past two

centuries.17 Again, neither study is of any pressing interest to the


currently practicing generative phonologist.

Finally, there are the two earlier-mentioned dissertations which

(in the fashion of my own study) relate Cuban linguistic data to broader

questions of theoretical interest. Hammond 1976, as noted, examines the

rapid speech rules seemingly distinct from ordinary phonological rules

through observations from MCS. And Guitart 1973 investigates a claim

that language-prt icular constraints (in this case phonetic neutraliza-

tion phenomena in ESH) are explained with a theory of generative phon-

ology amended to incorporate the SPE notion of Markedness. Data and

analyses presented in these final two works are somewhat crucial to

my own assessments of Cuban Spanish and this is a point which will be

pressed in detail with Chapter Five. Specifically, Guitart's obser-

vation of widespread velarization in Cuban Spanish and Hammond's first

account of a spreading spirantization process and the possible implica-

tions in terms of restructuring of underlying inventories are both in-

tegral (albeit in a largely negative sense) to my own contrary argument

for a Natural Phonology analysis of MCS and my thesis that innate pro-

cesses and not learned rules will account most insightfully for the

native phonology and borrowing phonology which together characterize

this special and emerging dialect.

Additional claims about MCS and the bulk of the data for this

study derive from my own field work in Miami and Gainesville carried

out over the eleven month period between January 1975 and November 1975.

The entire corpus of recorded material consists of twenty-seven tapes

of native informants and Spanish-language radio programming and totals

approximately fifteen hours of recorded speech. All taping was done

with high-quality Memorex cassette tapes on a Lloyds cassette-style


recorder; however for the purposes of analysis these tapes were either

played back through high-quality stereo speakers or listened to through

a set of stereo headphones. With the informant tapes, both types of

analysis were made. Twelve primary informants were taped in formal or

semi-formal interview sessions in either Gainesville or Miami, though

all had lived in Miami for a minimum of five years, used Spanish as their

dominant language (that is, at least in home environments), and resided

no extensive periods outside of Cuba and Miami. Several additional

informants were also recorded in the course of gathering data for this

study, but as these either had difficulty with the presented question-

naire, showed almost no knowledge of anglicismos current in Miami style

Spanish, or had resided in Miami for only several months, their tapes

were discarded for the purposes of analyzing this investigation.

These formal-interview tapes were supplemented with approximately

ten hours of additional cassette recordings of Miami Cuban radio pro-

gramming consisting largely of talk-shows, newscasts, and commercial

messages, with almost all recorded music having been eliminated. The

original purpose of this radio taping was to gain some general familiar-

ity with frequent and phonologically-interesting borrowings as well as

to solicit items for inclusion on the linguistic questionnaire. Radio

tapes eventually proved a fruitful source of loanword pronunciations,

however, and the use of this corpus in the phonological analysis of

borrowings in Chapter Seven seems fully justified in light of our

assumption that the predominant source of phonetic borrowings is the

first renderings by largely bilingual speakers who are in turn imitated

by the monolingual members of the speech community. Also, as a bulk

of English borrowings into MCS are place names and product names, a


fecund source for determining the original phonological shape in the

target language (Spanish) for the borrowed word (a crucial factor in

analyzing the course followed toward nativization) is radio commercials

and radio news broadcasting. Additional discussion of my field work

methods and a list of informants and phonetic forms is provided in

the Appendix following this study.

The questionnaire (also see Appendix) used was one designed solely

for this project and contains all the following types of closely related

interview items: (1) sentences featuring interspersed English borrowings

(downtown, bisi "busy", appointment, hacer espi "to hurry", and the like)

and to be repeated by the informant after an initial repetition by the

interviewer; (2) printed lists containing similar borrowings from which

informants were asked to repeat (i.e. read) each word and briefly re-

late what it meant; (3) numerous sentences of the "fill-in-the-blank"

variety to be repeated, with the precise word required by the blank

often being largely irrelevant as the items were frequently designed

to solicit pronunciation of another word or words in the sentence; and

(4) several questions designed to elicit "free discussion" and several

printed illustrations (newspaper photos and Norman Rockwell prints) for

the same purpose of eliciting free and informal conversations. Each

interview was carried out in Spanish and lasted only about thirty

minutes. Admittedly, having the informant repeat items already first

pronounced by the investigator or read from printed forms often leads

directly to questionable results; at the same time, no other approach

seemed practical or offered any real hope of eliciting certain desired

responses. The shortcomings and built-in inaccuracies of this method-

ology were kept clearly in view when selecting genuine loanword pro-


nunciations for inclusion within the corpus of transcribed borrowings

which comprises the results of this investigation.

It must be emphasized here initially that the discussion of Miami

Spanish featured in Part Three of this dissertation is not in any sense

intended as complete listings or a complete phonological description

of all English borrowings (nor even a representative percentage of them)

current in this dialect. There is, of course, no fixed number of loan-

words in any dialect where such borrowing remains an active and ongoing

process of linguistic change. I have caken as my subject just those

processes apparent in borrowings which provide insights into the workings

of loanword phonology and justifications for a system of Natural Phon-

ology approximating that outlined first in Chapter Two through Chapter

Four. Drawing heavily upon the list of loanword proposals systematized

by Lovins (see pages 21 to 23 above) as well as other hypotheses which

would seem to arise inevitably from the theory offered by Stampe, I

have sought out evidence in Cuban Spanish bearing on such fundamental

yet still unverified claims as those in the abbreviated list of loanword

hypotheses on pages 8 and 9 of this chapter. Three in particular give

shape to Chapter Seven and merit special repetition here:

Hypothesis I: The propensity of a language for borrowing words

is to a large extent determined by what shapes are permissible at its

surface level -- that is, it is generally late syntagmatic processes

and not underlying restrictions paradigmaticc processes or morphological

rules) which are relevant in determine what can be borrowed and which

operate on lexical items once they are adopted by a target (borrowing)

language system.

Hypothesis II; Loanwords have their underlying representations


established as phonemic representations and are nativized through

application of allophonic processes, facts which lend credence to the

proposals for a "natural phonemic" level.

Hypothesis III: Only the "living" processes, which are fully pro-

ductive in the grammar and will admit no exceptions, and not the "dead"

and unproductive rules, function in initially establishing a loanword in

the target language.

A goal of this dissertation, then, is ascertaining whether proposals

of this type (introduced with Stampe 1969 and first studied formally in

Ohso 1971 while not being explicitly codified until the works by Lovins)

withstand perceptive analyses of borrowings into some language other

than Japanese and English. And also, whether or not the evidence from

loanword phonology is to any important degree a confirmation of more

general principles advocated by the Natural Phonology espoused by



Some clarification would seem to be needed here, given that
Lovins's dissertation, as far as it goes, offers an excellent intro-
duction to the various approaches to problems in loanword phonology,
including approaches within Natural Phonology (esp. in Section
And Ohso's earlier Master's thesis is also a pioneer effort in Natural
Loanword Phonology which also repays most careful study.
Lovins (1973) and Ohso (1971) both examine English and (in Lovins)
other Western borrowings in Japanese, a data source which has been the
only language-contact situation so far to attract the handful of lin-
guists advocating Stampe's approach to loanwords. (This is in large
part no doubt due to the fact that James McCawley, the mentor of both
Stampe and Lovins, is the prominent student of Japanese among the
generative phonologists, as well as the fact that Ohso is herself a
native Japanese.)
Yet for all their many merits both these works are only reason-
able first approximations at treating loan phonology data within Stampe's
methods. Ohso gives sparse treatment to the theoretical implications of
her Japanese forms. Explicit discussion of Natural Phonology and Natural
Loanword Phonology in Lovins 1973 is likewise restricted to three brief
sections of Chapter Two (pages 25 to 37), and this in a thesis which
on the whole explores numerous approaches and several plausible solutions
to loanword phenomena. Fuller theoretical discussion of loanword phon-
ology in the light of Natural Phonology is to be found in Lovins 1974a
and Lovins 1974b.
As long as Natural Phonology remains in its infant stages, it
is of course not surprising that pioneer works should reveal such clear
limitations. The considerable potential of efforts in Natural Loanword
Phonology becomes somewhat more apparent in Lovins's more recent work:
viz. in Lovins 1974a, which explores the "stratification" of the lexicon
into "native" and "non-native" subcomponents from the standpoint of
Natural Phonology (with a conclusion that lexical representations have
stratal attributes as a direct consequence of which rules or processes
have already applied to them, an interpretation granting to lexical
strata "a rather different mode of existence"); and in Lovins 1974b,
which presents for tentative consideration the hypotheses outlined
above beginning on page 21 and subjected to some further analyses below
in Chapter Seven. The sum effect, however, of these earliest attempts
is only to emphasize the obvious fact that no extensive or even adequate
study of loanword phonology as parallel to Natural Phonology is yet

2For example, one widely cited and otherwise quite exceptional
paper (Kornfeld 1971) has summarized recent tenable theoretical positions
in child phonology without even once mentioning Stampe's hypotheses
among them.

One of the difficulties with Stampe's present theory is the
notion of the form in which the child inherits an innate system. This
issue must for the time being remain unresolved. Our only evidence
is that demonstrating the unalterable fact of innate processes.

Rhodes (1972) argues differently concerning restrictions on
underlying forms in Natural Phonology. It is Rhodes's conclusion that
such restrictions are not processes (actual substitutions as opposed
to traditional morpheme structure constraints) simply because "the lack
of ordering relationships between RUFs (restrictions on underlying form
-- PCB) suggests that RUFs are, in fact, simply conditions, all applying
at the same level in the derivation, i.e. in underlying forms" and
therefore "RUFs are MS conditions" (1972, pages 555-56). But the
difficulty is that Rhodes considers only one type of process as real
-- the context-sensitive -- whereas it is paradigmatic natural processes
which Stampe has proposed as being some kind of "filter" or as restric-
tions on the lexicon. This confusion in types is taken up more exten-
sively in Chapter Two, especially in Section 2.2 below.

Stampe acknowledges that this position is not at all unique and
that Passy, for example, expressed such a view at least as early as 1890.
See, e.g., Stampe 1969, page 448.
Lovins gives an example confirming this claim that allophonic
distribution is relevant to characterizing the "closest sound" in the
target language. Though phonetic [s] occurs in Japanese, any foreign
[s] is rendered as phonetic [s] only when it occurs in a context which
permits the phonetic [s] in Japanese.

However, Ohso notes that dominant (context-free in her termin-
ology) processes may indeed influence heavily the perception of foreign
sounds, and their production as well (see Ohso 1971, page 42; Lovins
1973, page 32). Ohso labels as "dominant" those processes which con-
strain underlying inventories of segments, and she apparently has in
view here only processes of the paradigmatic type ("X ->Y eliminating
X from the lexicon"). Throughout the discussion which follows and
in the model for Natural Phonology given in Chapter Four, I employ the
label "dominant process" normally in the same sense, though this may be
some distortion of what Stampe originally intended. Stampe has empha-
sized (personal communication) his own designation of "dominant" refers
to any process which applies earlier than a second one and therefore
establishes the environment for the second process. The context-free
processes are normally always prior to contrary context-sensitive pro-
cesses which counteract their effects (e.g. general vowel denasalization
must always precede contextual vowel nasalization in English). But the
syntagmatic processes might also dominate over other later and similar
syntagmatic processes, given this original interpretation of "dominant"

Lovins mentions at several junctures (e.g. see Lovins 1973,
page 27, or Lovins 1974b, page 249, note 9) that this phonemic level
of Natural Phonology is not a "uniformly positioned" level of under-
lying representation. This is a point about which Stampe himself is

not always especially clear -- an oversight which leads to most of the
difficulties of interpretation discussed in Chapter Two. Lovins's own
definition of the phonemic level in Natural Phonology is worth quoting
more fully:

This "level," in Stampe's view, doesn't really exist per se,
being peculiar to each process and each morpheme: it is no deeper
in each case than is needed to handle allophonic (contextual)
variation. It is therefore defined solely by the interplay of
context-sensitive and context-free processes, the former allowing
(when "allophonic") surface segments that have previously been
disallowed by the latter (Stampe 1972b) (1974b, page 249).

Lovins points out that similar suggestions about a possible
phonemic level have recently been given in Schane 1971, Wang 1968, and
throughout the works of Kiparsky. I would suggest that Lovins is guilty
here of the same misinterpretation as Rhodes (see the discussion in
Chapter Two); it is the interplay of allophonic and morphophonemic
processes, not context-sensitive and context-free processes, that is
operative in establishing a phonemic representation. Once this is seen
we are led to the view (presented in Section 3.1.2) that Stampe has
something much more heretical (in light of the standard generative
theory) in mind when suggesting this view of a speaker's notions of
phonemic representation.

Stoel claims:

Stampe is not explicit about the possible number and range of
types of "processes" which might be found in phonological acquis-
ition. Consequently his theory, as described in his own works, is
virtually untestable.........Jakobson, in contrast, makes precise
probabilistic predictions on the order of acquisition of phonemic
oppositions.........Jakobson's theory is more suitable for testing
than Stampe's (which makes it, in some sense, a better theory)
(1974, page 12).

Ohala also complains about testability in Stampe's position on
child language:

Stampe suggests that children are born with a set of universal
phonological processes in their head (sic) which tell them, among
other things, to produce their stops without voicing, to have no
consonant clusters, to drop h's, etc. The children learn how to
speak their particular language by deleting or otherwise modify-
ing this innate list of directions on how to talk. This is cer-
tainly possible -- I wouldn't say plausible, though. But this
hypothesis is too powerful: there is almost no data that couldn't
be explained in this way. There is then practically no way of
testing the hypothesis, in which case it is a toss-up whether
we accept this untestable hypothesis or some other untestable
hypothesis, e.g., "things are the way they are because God wants
them that way." Frankly, if it is just between these two, I'd
take the latter hypothesis because it even "explains" things
outside phonology (1974, page 268).

This view is modified, however, by the following retraction:

My mistake, apparently, was to take Stampe's "innate phonological
processes" to be software constraints mentalisticc controls -- PCB).
Stampe assures me that they are the constraints inherent in the
human body, i.e., hardware constraints. In this case I have no
argument with Stampe's theory -- indeed, it coincides with my own
view of the subject (1974, page 271).

Ohala is still troubled, however, by how "hardware constraints can be
'suppressed' or 'ordered'" and he questions what it would mean "to
'order' one innate physiological constraint after another" (1974, page
271). An answer in part is that such restrictions on pronounceability
are interpreted to be mentalistic as well as physiological, a point
which Stampe has spent considerable time defending in the opening pages
of his own dissertation (Stampe 1973a).

10Of course it is true that Natural Phonology to date has had
almost nothing to say about a third rule type also relegated in SPE to
the nebulous category of "phonological" component: morphological or
morpholexical (Anderson's term) rules. For possible definitions of
such rules see e.g. Halle 1973; Aronoff 1976; Skousen 1972, 1973,.1974;
Hooper 1973; Anderson 1974, 1975; and Cearly 1974. Chapter Four of
this present dissertation is, as far as I can tell, the first attempt
to discuss morphological rules, however superficially, within a treatise
on Natural Phonology.

1Terrell makes a parallel observation in this regard when he
comments that

The preoccupation of standard generative phonological theories
with "competence" contribute (sic) to the difficulty which studies
based on this model have had in coming to grips with an overall
phonological theory with explanatory adequacy. Explanation has
been, in general, relegated entirely to the formal apparatus of
a generative grammar with an associated "evaluation metric".....
Traditionally, however, questions of perception, production, and
acquisition, i.e., performance, have been considered as explanatory
bases for phonological processes (1974b, page 1).
1Skousen's approach is discussed at length in Clayton 1974,
Chapter One and Chapter Four. Clayton's dissertation is itself one
len-thy consideration of proposals in the recent literature to limit
phonology to "productive," phonetically-plausible rules and of some
speculations about the theoretical implications inherent in such

1Saciuk's approach to the "stratification" of the lexicon
has received criticisms for its excessive "abstractness" from some
recent sources. Wanner (1972) presses the conclusion that Saciuk's
scheme for "lexical stratification" can not be part of the phonological
description of any language, since it violates conditions on empirical
falsifiability. The essence of the argument is captured in the two
following paragraphs:

Another detrimental implication of the lexical stratification
hypothesis has to do with the claim it seems to make about the
status of its generalizations within the grammar: It is inconse-
quential for the speaker's knowledge/usage of his language whether
the underlying phonological representations are abstract or not,
i.e. whether he has analyzed the data along the lines of lexical
stratification or whether he arrived at a shallow phonemicization.
Lexical stratification therefore does not make any true predic-
tions, it merely states lexical relations on a static basis (1972,
page 12).

It turns out however that in a description making use of
lexical stratification exactly these erratic gaps in the lexicon
of a particular speaker will be responsible for the difference
between the speaker's having an "abstract" vs. "concrete" under-
lying representation for an item such as leche. This means that
there is no further content associated with the notion of under-
lying representation if it is the case that it may vary extensive-
ly and radically across idiolects. Since the difference in under-
lying representation is not correlated with any surface difference
of comparable impact on the overall grammar the conclusion is in-
evitable that the concept of lexical stratification rests on non-
empirical, arbitrary foundations (1972, page 13).

A second criticism more relevant to present concerns of our own
study comes from Lovins (1973, page 156). Lovins argues that "if
any of Saciuk's surface forms were taken as underlying (phonemic),
which they apparently are in relation to synchronic processes, it
would be more obvious that underlying representations have stratal
attributes only as a function of what rules and processes have already
applied to them" (page 156). The analysis by Lovins suggesting that
"for forms containing no exceptions to current phonetic processes,
it is precisely the phonemic level at which 'degree of nativeness' is
intuited" will come under more careful scrutiny as soon as we turn
to loanword phonology proper in Chapter Six and Chapter Seven.
A pressing need also will be for research like that now being
carried out by Reimold (1974a, 1974b). This is the effort to define
in formal terms what qualifies as a natural phonological rule (i.e.
Stampe's processes). Where Stampe is concerned with explaining the
nature of substitutions which children and later adult speakers will
actually make in arriving at utterances (that is, with the question
of "performance"), Reimold (like most generative phonologists) is
more interested in abstract properties of formal grammars. Reimold's
purpose is to set up structural definitions for certain natural rule
types which will also serve to account for phonological processes
in a more explanatory fashion (1974b, page 2). While according to
the standard view (viz. that of Chomsky and Halle, and Schane) natural
rules are presumably "simpler" rules, Reimold proposes to define
meticulously natural rules in strictly structural terms, so that it
will be apparent from the form alone what type of process (e.g. one
of weakening, or assimilation, etc.) a rule represents.

Although Guitart claims to characterize the Educated Spanish
of Havana (Cuban but not Americanized-Cuban Spanish) which is his own
idiolect, and though Hammond and I treat exclusively Miami Cuban Spanish,
we are in the long run discussing intimately related dialects about
which many assertions (e.g. Guitart's claim on the velarization of the
nasals and obstruents) would be equally applicable. I draw this con-
clusion largely from Cuitart's admissions that his own informants are
himself and selected personal acquaintances about whom he has commented
as follows: "Although the writer and other Cubans known to him who
speak this variety (i.e. ESH -- PCB) have been living in the United
States for at least five years and are more or less fluent in English,
none shows in his Spanish any noticeable phonological influence from
English" (1973, page 69, note 1). In other words, Guitart's informants
(to the extent that he uses any) are identical to mine and to Hammond's
(at least to speakers from our own samples who are university-educated);
that is, they are Cuban who have been living in the United States for
five years or more and who maintain Spanish as their first language yet
are somewhat fluent in English. It is on the basis of this similarity
in our samples that I claim to be justified in advancing criticisms
(see Chapter Five) of Guitart's few conclusions about ESH and about
educated Cuban Spanish in general.
As noted in the text, Cuban informants recorded in Miami also
serve as basis for additional studies: cf. Terrell 1974a, 1974b, 1975a,
1975b, 1975c, 1975d; Vallejo-Claros 1970; and Haden and Matluck 1973,
1974. What is not often pointed out is that all these studies draw
upon a single set of several hundred taped interviews recorded during
the period 1968 to 1970 by Bernardo Vallejo-Claros under auspices of
the aforementioned "Coordinated Study of the Linguistic Norms of the
Principal Cities of Ibero-America and the Iberian Peninsula" (directed
by Joseph H. Matluck and Humberto Lopez-Morales. A full description
of these tapes is given in the Introduction to Haden and Matluck 1974.
Although the number of tapes employed in each of the individual studies
mentioned above varies greatly, in all cases the informants were natives
of Havana and none had been in the United States (Miami) more than three
months at the time they were interviewed by Vallejo-Claros. In sum
total, the project taped over 400 hours of Havana speech in a series
of 454 interviews utilizing 750 informants, but all these "interviews"
were largely free and unrestricted dialogues between two informants
or the interviewer and a single informant and "secret" recordings of
spontaneous dialogue or of more formal lectures or speeches, etc.

1Like Guitart, both Lamb and Sosa claim to study the Spanish
of Havana but base their analyses on interviews with informants who have
been residents in the United States for at least a number of months and
often a number of years.

Part Two

The Processes of Natural Phonology



Assuming that there is some phonological level more abstract
than what has been termed the systematic phonetic level, what
are its properties and its relationship to the surface phonetic
level? This question, the most basic of phonology, is still
without a satisfactory answer.
-- Mary Louise Clayton

2.1 Preliminary Remarks

Stampe's conception of Natural Phonology potentially relieves the

contemporary theory of phonology of some of its most celebrated pseudo

issues. It seems to provide for the resolution of even such particular-

ly thorny issues as rule ordering (e.g. in so far as it sorts out se-

quential and linear properties peculiar to both "rules" and "natural

processes") and the "abstractness" controversy (see Kiparsky 1968a,

Crothers 1973, Schane 1974, etc.) in generative phonology. In the

introductory chapter we have surveyed some of these characteristics

of Natural Phonology and delimited the field of this present study.

In this second chapter and the one following I will now relate in more

detail precisely how the Natural Phonology of Stampe resolves contro-

versial issues within the generative phonology inaugurated by Chomsky

and Halle.1

As suggested in Section 1.2, unique to Stampe's model are a

concept of "natural phonemic" representations less abstract than

underlying forms in standard generative phonology and a related dis-

tinction between the acquired "rules" and innate "natural processes"



of a phonological derivation. By making exactly this distinction

between learned rules (the limited case which is imposed by each

language on its speakers in the sense that all rules are language-

specific) and inherited rules or "processes" (the more frequent case

which is imposed by the speaker upon his language in the sense that

all processes have a universal physiological or mentalistic motivation),

Stampe's version of phonology necessitates a considerable reassessment

in generative-transformationalist thinking. It is not quite possible

to ignore any longer, for example, that the practice in generative

phonology of lumping morphophonemic processes and phonetic processes

together as a single set of ordered rules has obscured a fundamental

distinction among the processes of phonology and misrepresented unique

teleologies of rules (Stampe 1973a, p. 43; Schane 1971; Anderson 1974,

1975). Yet the scarcity of published versions of Stampe's papers and

resulting confusion over the precise meanings of several of his claims

(e.g. the issue of what is actually "innate" in language and what is

ultimately "natural") has deprived this alternative model of consider-

ably greater impact to date and therefore motivates full well the

present chapters.

An observation of opening remarks in Section 1.1 was precisely

this scarcity of Stampe's published writings. Stampe's formulation of

Natural Phonology is mostly restricted to papers read to the Chicago

Linguistic Society, and though several of these talks since 1969 have

reached print others before 1968 have never been readily available.

Rhodes's articles (1972, 1973), on the other hand, stand as the most

elaborate attempt prior to the 1974 Papers from the Parasession on

Natural Phonology (cf. Bruck et al. 1974) at an objective critique and

an explication of Stampe. Yet unfortunately for natural phonologists,

Rhodes's papers with all their insights contain some troublesome mis-

conceptions and inaccuracies which so far have completely escaped any

commentary. With the appearance in 1973 of Stampe's own frequently

delayed dissertation (which aims to answer "shortcomings brilliantly

enumerated by Rhodes 1972 which natural phonology seems to exhibit if

the role of contrary innate processes in determining underlying repre-

sentation is not correctly understood," see Stampe 1973b, p. 1) we now

have a much fuller text from which to assess claims in Natural Phonology,

and it appears inevitable that numerous improved treatments will now

eventually be forthcoming. However, since we currently still lack

such treatments, this second chapter aims to contribute to a more

extensive debate of Stampe's principles and to a more thoroughgoing

review of Stampe's position than that found in previous works. To

achieve anything measuring close to this goal it will be necessary to

address here precisely those issues which are at present the most lively

issues of Natural Phonology and of generative phonology, as well as of

linguistic theory at large. These are the familiar impenetrable ques-

tions of:

(1) How are rules'ordered in grammars of natural languages?


(2) What precisely constitutes the underlying representations of

grammatical formatives?

In short, what are the degrees of abstractness admissible in phono-

logical grammars, and what are the numbers and kinds of phonological

substitutions that speakers make?

In commenting on these issues my focus will be principally on

the interpretation of Stampe by Rhodes and on Miller's subsequent

arguments (Miller 1975) against Rhodes's own variations of Natural

Phonology, as well as on some central features (the Phonemic Hypothesis

and evidence for context-free processes) which I take to be the most

often misinterpreted facets of Stampe's otherwise fully explicit


2.2 Conceptions of Rules and Natural Processes

Particularly unsettled at this point is Stampe's division between

rules and natural processes. As a case in point, we note that Rhodes

(1973) gives one sample derivation in terms of Natural Phonology which

suggests that a "natural phonemic level" separates all "rules", which

apply before, from the assorted processes applying after. This is

reminiscent (Rhodes's own statements to the contrary notwithstanding)

of a classical phonemic level which divides morphophonemic rules which

result in surface constrasts from the phonetic rules which do not

(cf. Schane 1971). But the proposal is not especially helpful in

elucidating Stampe's own suggestion: i.e. that it appears to be a

level of "short-term memory storage" (having its reality as the locus

of "slips of the tongue") which is the expected point of division

between what are rules and what are processes (Stampe 1973a, p. 44).

Stampe's more specific claim is that "phonological processes apply

after slips occur, and therefore must also apply in the processing

of individual segments" (1973a, p. 44); and therefore, by implication,

the contention is that rules (irrelevant to such processing) apply

only before such slips. Rhodes's division of the rules, when coupled

with Stampe's suggestion, makes the strong claim seemingly amenable

to empirical evidence that "slips" occur at the point of phonemic

analysis; yet it does so without making altogether clear what Stampe's

concept of the phoneme and its representation in generative grammar

actually is. Rhodes seems to assume, then, that since processes are

sensitive to such slips while rules are not, it follows that processes

are all of a single similar kind and all act upon (i.e. follow in the

derivation) what Stampe would call underlying (phonemic) forms. Much

of this particular chapter is devoted to showing the recklessness of

any such proposals.

Miller (1975) claims to refute Rhodes's tentative model for Natural

Phonology on the basis of a single example taken from ancient Greek of

an "unnatural" substitution (a "rule" of initial r-aspiration) which

must apply down among the late natural processes, given its status as

a completely automatic and therefore late (low level) process. Miller's

complaint is not with Stampe's apparent resurrection of a phonemic level,

which would be expected to be a popular objection among most generative

phonologists, but with Rhodes's quite arbitrary divisions between rules

and processes. Very loosely Miller's criticisms seem justified, in so

far as Rhodes's division and descriptions of rules and processes don't

match with Stampe's. Yet one of the points I wish to emphasize here is

that Rhodes's portrayal of Natural Phonology inaccurately represents

details of the theory for reasons that have nothing to do with the

example put forth by Miller. Also, Miller's example of r-aspiration

in Greek is subject of an analysis which destroys any and all contention

that it disproves the general claim that rules precede processes and

that processes are "automatic" and hence the "natural" substitutions

in the grammar.


Actually, three separate and crucial issues appear to be raised

with these two papers (Rhodes 1973 and Miller 1975) and might be best

summarized as follows:

(3) Rhodes claims only learned rules occur before the phonemic

level and only innate processes apply after such a level.

(4) Rhodes additionally claims some rules (preceding the phonemic

level) display characteristics of processes and therefore are to

be classified as "natural rules," although rules (learned behavior)

and not processes is most assuredly what they are.

(5) Miller counterclaims there are some automatic late processes

(after the phonemic level) that are learned and therefore a type

of "unnatural" (as opposed to "natural" in Stampe's sense), though

nonetheless automatic process.

Therefore, whereas Rhodes suggests rules functioning something like

natural processes (e.g. Turkish Vowel Harmony, taken up below) which

results in both natural rules and natural processes, Miller's position

is that there can be rules (or at least "unnatural" processes) ordered

among the automatic "phonetic" substitutions and thus below the phonemic

representation. I find it somewhat significant, however, that neither

seems overly concerned with investigating what actually it is that

corresponds to "phonemic representation" in these sample derivations

they present; that is, neither paper gives any evidence such as that

found in Stampe's dissertation for how speakers must determine what

would constitute true underlying sequences from which they derive (at

any rate of speed or in any style) strings like "Is pot good?" (which

is Miller's example) or a standard pronunciation of e.g. "electri[s]ity"

(Rhodes's sample derivation). Or, for that matter, if the speaker


makes any phonemic analysis at all. This last question, I would want

to argue -- not merely the orderings of rules versus processes, which

may soon become a non-issue at any rate, given the shakey status of the

present assumptions about sequential orderings of rules -- is the single

relevant question about any system of Natural Phonology.

I will proceed to demonstrate here that all three issues labelled

(3) through (5) are at odds with Stampe's repeated position and hence

not appropriate claims about Natural Phonology. Whether or not they are

to be taken as correct statements about language in general is an issue

we will studiously avoid at present, with evidence favoring Stampe's

model being postponed for later chapters. I begin instead from an

assumption that the argument of this chapter -- viz. that both Rhodes

and Miller to various degrees misrepresent Stampe and Natural Phonology

-- is prerequisite to any further discussion of the natural phonologist's

model as it is described and in some facets only implied in Stampe's

own papers.

In this section I deal exclusively, then, with oversights in Rhodes's

and Miller's discussions of Stampe. A further treatment of this topic

in Section 2.3 dismisses as well the superficial arguments in the recent

literature against the phonemic level as a basic level of linguistic

analysis and against the reality of Stampe's governing distinction of

paradigmaticc versus syntagmatic" among natural processes. Other issues

are deferred until Chapter Three and Chapter Four, where I give fuller

explications of what I assume Stampe has intended for a proper model

of the phonological component within a workable theory of natural


As a point of departure, consider the following responses to (3)


through (5), which we will label here (3') through (5') and upon which

we will elaborate throughout the remainder of this chapter.

(3') The claim that all processes follow (are ordered subsequent

to) Stampe's phonemic "level" ignores Stampe's own announced position

that an underlying representation in Natural Phonology is "phonemic"

only in the sense that it is no deeper than what is necessary to

account for allophonic variations -- Rhodes's claim therefore does

not seem to allow for the deeper morphophonemic processes and rejects

altogether the existence of paradigmatic processes, the latter

normally applying in unrestricted fashion to govern exclusively

the stored or lexical representations.

(4') The proposal for "natural rules" to be contrasted with the

"natural processes" results at least in part from failure to take

into consideration a full set of defining characteristics of innate

processes, like that presented e.g. in a middle section of this

current chapter and first tentatively put forth as an explicit list

in Bjarkman 1975.

(5') Miller's example from classical Greek will not in point of

fact stand up as a valid counterexample, since when the correct

historical analysis is supplied the rule in question appears to

meet all obvious features of a natural process as opposed to a

learned rule -- yet even the proven existence of some undisputed

case of the type Miller has in mind would not do irreparable damage

to the claim of Natural Phonology, since Stampe does admit (see

Stampe 1973a, page 66) the possibility of rare and unnatural cases

(and I speculate that these cases might well all involve what are

clearly "morphological" rules) where rules might indeed apply after


natural processes.

The model Rhodes proposes might conveniently be labelled as the

Sequential Rules-Processes Model (employing my own label) and can be

outlined as follows (with capitals indicating levels of representation

having some assumed psychological reality and the smaller case letters

indicating types of intervening processes).



where: UNDERLYING FORM = what
generative phonology calls
Systematic Phonemic Form

what Stampe commonly refers
to as the "underlying level"
but what Rhodes argues is not
a Classical Phonemic level

Implicit here is what Rhodes has interpreted as Stampe's claim

that no "rules" are ordered down among the natural processes (for some

explanation of Stampe's account see Section 3.1.2), as well as an even

stronger assumption that all rules precede and all processes follow a

phonemic level of representation. With this model it must be concluded

that a phonemic representation exists as an autonomous level within a

rule a

(rules and "natural" rules)

rule z
process a

(natural processes)

process z


generative phonology (i.e. there is a line in a generative derivation

exclusively equivalent to a phonemic representation), that such a level

functions to distinguish rules from processes, and that these latter

represent two distinct kinds of substitutions in natural language. The

final claim, of course, is unquestionably a claim made by Stampe.

It is useful to review here these assertions by natural phonologists

that two distinct kinds of substitutions operate in natural languages.

The first and demonstratably most numerous type (subject to the closest

scrutiny in literature on Natural Phonology) are the innate processes

(often labelled simply the NPs) which have a mentalistic as well as a

physiological motivation. Phonological processes function to merge

the potential phonological opposition which are characteristic of

speech into a single member of the opposition which most conforms to

restrictions on the human speech capacity. That is, into the member

(voiceless as opposed to voiced obstruents in final position, for

example) which is easiest to articulate. These opposition are the

contradictory sets which reflect the conflicting phonetic restrictions

underlying all utterances. As Stampe expresses it, processes attack

phonetic difficulties in appropriate and natural ways, but ways which

often are in conflict with other potential processes (Stampe 1973a,

page 41). Speakers will quite "naturally" devoice obstruents, since

their oral constriction impedes an airflow vital to voicing. On the

other hand, in voicing environments, say between vowels, a voicing

process becomes the "easier" alternative by assimilation. We intro-

duced above Stampe's assumption that a speaker can resolve conflicting

processes through the suppression, limitation (a context-sensitive

subcase of suppression), and reordering of these processes (Stampe


1969). In its most primitive state, when first exposed in the child,

an innate phonological system will display the full array of such

restrictions on speech capacity; it is the complete set of unordered

and as yet still altogether unlimited phonological processes. The

task of the child acquiring a language is, then, to make revisions in

all aspects of this system which are at first separating child-like

pronunciations from the standard of the adult pronunciations (Stampe

1969). It is a common observation already noted from the literature

on child language that many children first produce the English words

"dog" and "doll" in identical fashion, as [da]. But once velars are

acquired, "dog" becomes [ga], with an apparent assimilation being made

to a deleted final velar. (That is, underlying /dzg/ becomes /dag/

by lowering, then /gag/ by velar assimilation of some type, and then

/ga/ by final velar deletion. ) Eventually velar deletion is fully

suppressed yielding finally the standard output of adult speech.7

Through use of the three mechanisms for resolving articulatory

conflicts, the child "learns" to pronounce new phonetic opposition

that are revisions in his "innate" phonological system (for many more

specific examples see Stampe 1969; 1973a, Chapter One). The mature

speech system maintains just those aspects of an innate system that

this mastering of adult pronunciations has left intact, along with

some learned substitutions which are the "rules," and those processes

surviving this more or less "unnatural" sort of limitation of the

innate tendencies toward physiological ease will determine what are

the pronounceable phonetic representations of any given language.

(A process of final obstruent devoicing survives in German dialects

but not in English.) Failure to limit processes in this way would


potentially reduce all utterances to the single lax stop and low

vowel [ba] with which the child begins. Thus Stampe and his students

observe that you in fact can't learn a language (that is, if you are

a child) only "by doing what comes naturally." Acquisition becomes

a matter of suppressing utterances as much as it is a matter of

acquiring articulations.

The second kind of substitution in natural languages is the

phonological "rule", which is acquired behavior and accounts for the

abstract alternations which often occur in strictly morphological or

even grammatical environments. An example appealed to by Rhodes is

the English rule of Velar Softening. This is an alternation of /k/

with /s/ and of /g/ with /1/ (/dg/) before the reflexes of front

vowels in words with Romance origin (e.g. electric versus electricity

and pedagogue versus pedagogy), an alternation for which the lexical

restrictions (it applies in electricity but not in persnickity) and

therefore the learned nature is patently obvious. Stampe (1973a,

page 45) significantly observes that children can not be assumed to

begin with pronunciations of words like "kitty" and "get" with /s/

and /j/ and later to limit these segments to only words which are

the Romance derivatives. Instead, some words are quite obviously

learned with /s/ and /j/ in the first place; yet, as Stampe again

observes, those forms which are learned from reading sometimes fail

to show the expected alternations, and in fact one at times will hear

"pedago[g]y" or "fun[g]i" or the like.

To return to Figure One, there are two reasons which should be

more or less immediately obvious for rejecting Rhodes's Sequential

Rules-Processes Model as the proper description of the type of phono-


logical component Stampe is suggesting. The model taken from the

sketch above (Figure One) is inaccurate and misleading, firstly because

it ignores altogether a distinction Stampe and the natural phonologists

make between two functionally separate classes of natural processes.

These are context-sensitive processes (hereafter also called "syntag-

matic") which operate to ease articulatory difficulties in sequences

of segments, and context-free processes (herafter paradigmaticc") which

clarify phonetic properties of the individual segment. A full delin-

eation of the types is given in Stampe 1973a, page 21, and in Donegan

Miller 1972a. As paradigmatic processes are predominantly those which

will govern underlying representations, Stampe assumes they are neces-

sarily ordered prior to the syntagmatic processes, which will govern

consistently surface representations.

It is evident from scanning that Rhodes's model has only one type

of process in view and this is of necessity (since they are found

following the "phonemic" forms) the syntagmatic variety. More mis-

leading still, Rhodes says nothing about what I would credit as being

Stampe's distinction between the morphophonemic syntagmatic processes,

which substitute segments not previously eliminated by some prior

paradigmatic process (i.e. those not impermissible in underlying form)

and thus in the traditional sense "neutralize" underlying distinctions,

and allophonic syntagmatic processes, which substitute just those sounds

which do not occur in underlying (phonemic) strings. It would seem

imperative, then, to take exception to one observation (1973, page

537) by Rhodes in the face of all others: this is a statement that

devoicing is a natural process par excellence, and therefore that forms

at the natural phonemic level do not show the effects of devoicing,


that bunt "colorful" is represented as /bunt/ (cf. bunte) and Bund

"association" is represented /bund/ (cf. Bfnde) even though they are

together pronounced identically as [b't ].

A number of points in our discussion below depend rather on the

alternative assumption that German final devoicing is a morphophonemic

(pre-phonemic) process and thus that German speakers take as phonemi-

cally distinct the two forms that Rhodes cites as having identical

"phonemic" representations, one fact by the way which accounts well

for Shibatani's discovery that speakers take just such a "rule" as a

functioning restriction on possible forms of their language (for discus-

sion see Hooper 1975, page 543, as well as Section 3.1.2 below). Such

distinction leads immediately to a conclusion that some processes do

indeed precede phonemic representation (the morphophonemic type supply-

ing features, derivable but already prevalent in the lexicon, at some

level speakers take as underlying) and others (the allophonic ones)

do not. And such a conclusion straight away invalidates a model

postulating an intermediate level which has only the learned rules

coming prior to it and all natural processes ordered after it.8

We observed above that Rhodes also argues for a category of

"natural rules" functioning above (ordered prior to) the natural

phonemic level in particular derivations, a category which consists

of some otherwise natural processes which are for some unaccountable

reason now operating as rules, among which they are as a result ordered.

He would place substitutions like Turkish Vowel Harmony into such a

category of rules: i.e. "even though the physiological motivation is

clear, nonetheless, in most instances, languages having vowel harmony

substitutions are either replete with exceptions (like Turkish), or


contain 'neutral' vowels which obscure the phsyiological motivation

(like Finnish, and Nez Perce), or are opaque and strangely constrained

(like Yokuts)" (1973, page 534). But since rules which insert epenthe-

tic vowels must apply prior to Vowel Harmony and therefore must also

be determined to be of the nature of rules not processes, and since

other cases of epenthic vowels are more clealry NPs, it is seemingly

necessary to conclude that at least some substitutions must function

randomly in given languages as both natural processes and natural-type


Here I would contend that Rhodes is simply guilty of some faulty

reasoning concerning teleologies of processes. His mistake seems to

be one of equating the naturalness (in the Stampean sense and not the

sense assumed by Schane and other generative phonologists) of a rule

with "exceptionlessness" of appearance across many diverse languages.

Although single processes have transparent physiological and mentalistic

motivations, the fact that universality of occurrence is no fast criterion

is made clear enough by Miller's case (below) of r-aspiration in Greek.

English Flap-deletion is a frequently-cited process in the literature

on Natural Phonology but not necessarily a widely-surfacing one among

the world's languages. More than likely, the answer to the paradox

of Vowel Harmony, a much more straightforward case than Miller's

aspiration example, is, I suspect, an obviously historical one. What

began as a natural process with all the earmarks of processes (the

occurrence in child speech across countless languages, the frequent

appearance in adult speech of many languages, and direct physiological

motivation) has, by a normal chain of events (see Stampe 1972 on the

English Vowel Shift), converted to the learned rule of Modern Turkish.


The case would in fact be parallel to that of the Vowel Shift of

English. It is some centuries since such a vowel shift can be viewed,

as Kiaprsky (1968a, page 11) erroneously claims, as an automatic and

low-level phonetic process; and yet the overly abstract nature of the

representations and complexity of derivations in Chomsky and Halle's

version of the vowel shift become apparent when one looks at the many

non-alternating lexical items of English (see Krohn 1974; examples are

the underlying representations for vowels in words like moss, lawn,

maudlin). What is a process and what is a rule is not, after all, so

difficult to determine once the criterion of function is made the

primary measure. Vowel Harmony, replete as it is with such numerous

exceptions, simply no longer appears to be taken as a restriction on

pronounceability by the contemporary speakers of Modern Turkish.

To recapitulate, I have been suggesting here that Rhodes's formal

schematization of Stampe's model is inadequate if only because of its

failure to acknowledge a distinction between functionally paradigmatic

and syntagmatic processes, as well as a further contrast in types

between morphophonemic or pre-phonemic and later allophonic processes.

When we turn next to Miller 1975 we discover an attack on Rhodes's

schematization which is based on somewhat less substantial grounds

than these, yet one nonetheless worth pursuing since it does reveal

further potential misconceptions about the structure of a proposed

Natural Phonology.

Miller also rejects the Sequential Rules-Processes Model for

enticing reasons, though in the act he challenges a tenet of Natural

Phonology which Rhodes with good reason has assumed to be a "given" of

the theory. This is an assumption that substitutions which are totally


automatic (i.e. exceptionless, "transparent," palpably a restriction

on what is pronounceable and thus difficult physiologically to overcome,

inviolate in the face of borrowings, "slips," and word games, etc.)

correspond to Stampe's set of natural processes and are the residues

of what is innate for the speaker and not simply learned arbitrarily,

because this or that peculiar language requires certain pronunciations

for certain contexts. Miller attempts to argue that "learned" sub-

stitutions can also be in this very same fashion altogether automatic


To illustrate that there are rules which apply below Rhodes's

natural phonemic level, Miller cites the "rule" of r-aspiration which

applies invariably whenever this segment is initial in ancient Greek.

That "invariably" may be an overstatement of the case is a plausibility

we will take up further below but one which need not detain us here.

There are superficial indications that this particular substitution

displays features of a highly unnatural and therefore presumably learned

type of substitution for speakers of ancient Greek. Some of the evi-

dence would be as follows, selecting here only those most salient points

of Miller's discussion.

Universally, aspiration is recognized (in the general work of phone-

ticians like Ladefoged, for example) to be rather infrequent on liquids

and nasals. In the specific case of Greek, presumably no sonorant was

aspirated in underlying representation and in output strings it is only

word-initial [r] that has an aspiration. (We are assuming here along

with Miller that orthographic rh represents a voiced segment that is

aspirated; that this is not in fact the case will be treated thoroughly

below.) Miller emphasizes that r with aspiration is increasingly unusual


in this case since it contrdicts the otherwise impeccable generalization

that no voiced segment of this language was aspirated, a fact unexcep-

tionally true of the total lexicon. While an aspirated sonorant would

seem to be a highly marked segment universally, early Greek had mh, lh,

rh, nh (which contrasted with m, 1, r, n) before all were quite predict-

ably lost, excepting the case with the initial r. rh instead became

generalized to all initial occurrences, those from sr as well as under-

lying r.

What is crucial is the ordering of this r-aspiration, which is

applicable following at least two natural processes of (1) aphaeresiss"

(an optional deletion in fast speech of word-initial short vowels after

long vowels and diphthongs) and (2) "ancephale" (an initial vowel syn-

cope), both of which operate to expose r to the aspiration rule (e.g.

erimazo "I go along" -> rhemazo; erot' "I ask" -- rhotJ, etc.). Miller

concludes that what is more appropriate than a model like that sketched

by Rhodes for encompassing such data is the alternative model below,

which we might conveniently label as the Alternate Outputs Model for

Natural Phonology.




rule a

(rules and some natural processes)

rule x -P

reductions x -

reductions y -

reductions z -


ideal output

- output x

> output y

-> output z

Wherever the derivation stops in the grammar sketched above, there

the automatic substitutions, both natural rules and unnatural rules,

are given a chance to apply. The alternate outputs are, then, all

possible outputs, depending for their differences on the degrees of

increasingly rapid and hence casual ("sloppy") speech. One example

would be Stampe's celebrated derivation of the phrase "divinity fudge"

(Stampe 1973a, page 59; Lee and Howard 1974) where more than a dozen

unique outputs rest on repeated applications of the syllabication,

flapping, vowel nasalization, and shortening processes of English.


That some forms in Stampe's derivation are marked starred forms which are

unpronounceable because obligatory substitutions have not yet been applied

is again a peripheral issue which will not sidetrack us here. We can

point to a parallel example as being Miller's own illustration of the

phrase "Is pot good?", which could be reduced as far as [spatgUd] follow-

ing non-automatic complete Cliticization and Copula Reduction, with the

application of Aspiration being lost due to the loss of environment for

this late process following applications of the above two non-automatic

processes. Since the late automatic substitutions (Aspiration, for

example) are necessarily given the chance to apply after we reach the

immediately pre-output stage (following non-automatic substitutions like

Cliticization and Copula Reduction), they patently must apply after a

natural phonemic level in the proposed Sequential Rules-Processes Model.

Miller's argument in short is, then, that (1) the natural phonemic level

as presented by Rhodes is fallacious, since we seem to have a case of

an acquired and unnatural rule in ancient Greek following the phonemic

level and ordered among late processes; and (2) the natural phonologist

(Stampe as well as Rhodes would have to be included) fails to distinguish

carefully enough between a block of immediately pre-output automatic

substitutions, whether natural or not, and those substitutions which are

more obviously physiological in aim. The main point of contention that

Miller raises, then, however indirectly it may address Stampe's original

proposals, is the potentially devastating issue that all automatic

processes do not per force seem to be in any necessary sense truly

natural. It is this contention that there may be unnatural automatic

processes that must eventually stand up against the weight of empirical



One difficulty, of course, is with the entirely singular nature of

Miller's example. No case built on a single language-peculiar example

could be accepted as being more than merely suggestive.

Admittedly, it is fundamental to question any formal organization

of the grammar into rigidly prescribed blocks of rules in which these

relative orderings mechanically will determine the possible types of

extant rule interactions. But Anderson (1974) has already raised this

identical issue much more eloquently and with considerable cross-language

evidence suggesting in effect that morpholexical, phonological, and

phonetic rules are generally but never quite unexceptionally ordered

as mutually exclusive sets. The evidence from exotic (Abkhaz, Luiseno,

Cebuano) as well as familiar (American English and Danish) languages

is that although Morpholexical/Phonological/Phonetic is the anticipated

ordering of rule types, it is repeatedly in those cases where the

transparency of some morpholexical rule is affected by the presence

of a phonological rule that an unnatural ordering is encountered.

Anderson nonetheless deduces a general principle: "the rules of the

grammar which fall into the three types delimited above, with their

distinct formal properties, are inextricably intertwined in the grammars

of natural languages" (1974, page 3). The point of Anderson's quite

judicious paper is precisely that any principle of ordering faces the

realization that natural languages display unnatural rules and orderings

as well as natural or expected ones.

Also weighing against Miller's argument is the contrary evidence

that r-aspiration, so called, is in fact upon closer inspection reduced

to a natural process and appears to be not at all a learned rule which

operated in the phonological grammar of ancient Greek. This is most


particularly evident once we bear in mind that the distinction between

"rule versus process" is not dependent exclusively on relative orderings

in a grammar but more on the identifiable teleologies of the two formal

types. In this light there is some indisputabt]. _%idence that initial

rh is product of a natural process. To accommodate r-aspiration as a

natural process in ancient Greek we need only to return to the standard

descriptions of the substitution in the accepted handbooks (a point which,

in all fairness, Miller was himself in part responsible for suggesting

to me). Perhaps also helpful will be some minor though necessary modi-

fications of Stampe's original conception of innate processes. Before

we take up such modifications, however, it is advisable to clarify still

further the respective teleologies of both processes and rules. It will

be such modifications and such clarifications which are not only the

major contribution of this present chapter, but ultimately they serve

as well as justification for such extensive discussion of the reputed

Greek r-aspiration example in the first place.

Above we have established tentatively that "rules" are the learned

substitutions. A further characteristic of this type is that they apply

to lexical representations to alter categorical values of distinctively

specified features (i.e. they are not static conditions or redundancy

statements governing the contents of the lexicon but rather have an

exclusively post-lexical function); and yet they are not palpable con-

straints on pronounceability and therefore are quite insensitive to such

performance errors as tongue slips.

Furthermore, rules are seemingly never optional ("electri[s]ity"

never alternates with "electri[k]ity" in any known form of dialect); they

make radical substitutions of segments; they have abstract (grammatical


or morphological but never purely phonetic) environments and they are

always context-conditioned; they are probably linearly as well as se-

quentially ordered (though this is the issue least likely to find easy

resolution in contemporary phonology); they quite obviously do not con-

vert through mechanisms of historical change to processes; and,.finally,

they are not productive in the language in the productive sense (i.e.

they do not apply to nativize loanwords, in a fashion elaborated in

Chapter Six and Chapter Seven).

Processes, by complete contrast, are an innate residue of the

primitive language system. They may in fact determine what is per-

missible in the lexicon and are in a very real sense restrictions on

lexical forms (as with the paradigmatic process which eliminates under-

lying nasal vowels from the segmental inventory of English or the

syntagmatic process which determines that all stops will be voiceless

after tautosyllabic s of English). Additionally, processes are con-

firmed restrictions on pronounceability which can not be violated even

by slips of the tongue (i.e. they reapply after such slips); they are

often optional in so far as they are affected by speed and the style

of speech; they make only quite minimal substitutions; they are man-

ifested as either context-free or context-sensitive (the first type

being generally ordered earlier); they have always an obvious phonetic

function, given that their purpose is to eliminate difficult phonetic

configurations, and what are deemed lexical or underlying restrictions

may reapply on surface output (e.g. a restriction like that on English

stops after s extends to phonetic output in rapid speech pronunciations

of phrases like "let's go" yielding [skow]); they are both sequential

and non-linear, as well as random and iterative, in their applications;


they can potentially develop into "dead" rules (two classic cases in

the literature being Turkish Vowel Harmony and English Vowel Shift as

earlier mentioned); and they do prove to be productive in a synchronic

version of the grammar. That processes are not restricted to any

linear ordering, but apply and reapply over and again quite randomly

each and every time those configurations reappear which they are aiming

to eliminate, has been seemingly demonstrated by Stampe's now popular

examples of fast speech pronunciations for phrases like "divinity fudge"

(1973a, page 55).10 This principle can without much difficulty be

assumed a well-established fact of the phonology -- that is, at least

for the distinctions set forth in this present chapter.

We will now consider somewhat further these several defining

characteristics of natural processes as I have outlined them here,

especially with reference to alleged Greek r-aspiration as one quite

exemplary instance. Since the intention here is a clarification of

Stampe's principles governing phonological processes, I will limit

consideration largely to Stampe's own few repeated examples involving

English forms and English rules.

2.2.1 Non-suspendability of natural processes

Even the optional variety of process peculiar to fast speech

pronunciations is most difficult to suspend and therefore sensitive

to spoonerisms or slips of the tongue (compare Fromkin 1971). Rules,

though obligatory (Velar Softening is not an "optional" English rule),

are easily enough suspended and have numerous evident exceptions. The

initial aspiration of stops in English, say, will unavoidably apply

to the results of spoonerisms like khatt/ step] kotch stape or perhaps

[that/ sk,eip] totch skape for [skat/ theip] scotch tape, and the


palatality of [k] will similarly be adjusted in line with its peculiar

environment (Stampe 1973a, page 44). Yet Velar Softening can quite

easily be withheld by speakers to allow the facetious pronunciation

of "electri[k]ity" or a normal pronunciation of the rare but certainly

not unheard of form "persnickity" (see also Hooper 1975, page 544).11

The complete exceptionlessness of the initial rl in ancient Greek would,

then, tend to argue most persuasively for its legitimate status as a

full-fledged natural process, insensitive to any notion of suspendability

attached to rules.

2.2.2 Optionality of natural processes

Processes, on the other hand, are often only optionally retained,

relative to the governing style and the speed of articulation. There

is nothing as all automatic about certain "sloppy speech" rules of

English: viz. aspiration of voiceless stops initially in the stressed
/ h h '
syllable (e.g. [prapLr] becomes [p raph r] "prepare"), optional consonant

unrelease after penultimate nasal consonants (e.g. [hAmp'] becomes

[hAm'] "hump" or [hAnt'] becomes [hAn?] "hunt"), or syllable internal

d-deletion (e.g. [Endgem] becomes [Engem] "endgame" or [kajndnas]

becomes [kajnnos] "kindness"), or any other of a number of related

processes surveyed in Hill 1973 which are facilitated but certainly

not dictated by physiological and articulatory pressures.

Rules apparently will never have this optional property. Once it

has been acquired, Velar Softening (adopting this standard example as

or prototype of the notion "rule") will apply for each speaker to

"electricity", and yet as we have noted it does have evident exceptions

in some other forms with exactly the same phonological characteristics

(i.e. "persnickity" and "lickity split"). We must maintain a careful


distinction here between what I am calling suspendability and what we

label optionality: Velar Softening may be suspended by conscious effort,

as it has no apparent irrepressible physiological motivation and rep-

resents no restriction on pronounceability (it is no more difficult for

one to say electricityy" than "persnickity"), but it is normally applic-

able irrespective of style or rate. The exceptionless, automatic nature

of Greek r-aspiration, which seems to relate it to physical restrictions

or at least presumed physical restrictions of the speech capacity, is

again on this count suggestive of its status as a restriction on what

is perceived to be pronounceable by speakers ol this language. That

this is not a real restriction on pronounceability, as we shall see

promptly, is apparently of little real consequence to speakers.

2.2.3 Extremity of natural processes

Processes make only minimal substitutions (change only a limited

few features) while rules make more radical substitutions (change a

greater number of distinctive features of any segment). That this is

in fact the case has not always been so readily apparent. Yet where

the substitutions that are fully qualified processes appear, on the

surface, to make more radical changes in segments, as in the example of

Joan Velten's drastic replacements of [1] with [z] in a much-cited

case from literature on child language (Stampe 1973a, pages 12 and 63),

the case turns out to be one of sequential applications of several

distinct substitutions (1 -r j -> 4 -- z), each in its turn actually

making only the most minimal potential change. Delateralization in

this case functions initially to eliminate a difficult tongue con-

figuration; Spirantization will increase or amplify the audibility

of [j] with a sibilant; and Depalatalization replaces fronted and


raised tongue posture with the neutral position of plain alveolars.

The functioning of individual processes makes obvious a physical and

mental restriction on speech mechanisms that no single and sweeping type

change between abstract elements might ever encompass.

2.2.4 Non-linearity of natural processes

Stampe contends that processes are both sequential (they are not

simultaneous) and yet non-linear (they may apply and reapply) in their

applications. They are never merely static conditions like the morpheme

structure constraints or mere formalisms like the redundancy rules but

are actual substitutions working upon real segments in the mental and

the physical performances of the normal speech act (Stampe 1973a, page

43). This is an issue which merits our much fuller attention.

Stampe's theory assumes not only that "systems of phonological

processes are real, that underlying and superficial representations of

utterances really exist, and that they are constrained and interrelated

by the actual agency of these processes" but also that processes "are

actual substitutions occurring in the performance (mental as well as

physical) of utterances" and thus able to provide not only analogical

but literal explanations of phonological events (1973a, pages 43-44).

But it is the argument that processes are non-linear that is most

controversial in Stampe's phonology.12 This argument is set forth

most convincingly in terms of evidence from the several casual speech

pronunciations of the phrase divinity fudge analyzed in Chapter Two

of Stampe's dissertation. The analysis depends on several successive

applications and reapplications of the three processes of Syllabication,

Flapping, and Flap-Deletion. It is reproduced below (Figure Three),

for the convenience of our discussion here, precisely as it is found


in Stampe's original chapter. Here the derivation is taken up at an

intermediate stage and starred forms represent forms which remain

unpronounceable until later obligatory substitutions have been applied.

Even further reductions than those shown here of course remain possible.






[davlnati (fAdg)]















where r = English flap

da.vln. a. ti

*do.vl. a.ti



*da.vI D.ti

da.vla. ti





da.vir. i




I = front unrounded high vowel

* = unpronounceable form (i.e.


[davlnati (fAdr )]

1. *da.vIn.at.i

2. *da.v r.r.i

3. da.vir.ar.i

4. ?da.vT.a.i

5. *da.vla.i

6. *da.vII.i

7. da.vI.i

8. *da.vii

9. da.vIi


all relevant processes have not

The point of the divinity fudge analysis is quite obviously that

processes are nowise restricted to linear order but will continue to

reapply each and every time an input configuration will arise. This

claim has predictably not gone undisputed, however, and one especially

controversial alternative analysis of this derivation is provided most

recently by Lee and Howard (1974). In this alternative treatment the

reasons for questioning Stampe's interpretation are taken to be something

like the following.

(a) Stampe's derivation of divinity fudge and his definition of

Syllabication (as a process) are apparently in contradiction. To quote

directly: "According to Stampe's own description of the Syllabication

process, we would expect the [t] to be attached to the preceding syllable,

since the following syllable is unstressed. Such a syllabication would

allow the derivation (2) above [cf. Figure Three -- PCB], in which there

are several fewer reoccurrences of processes. Stampe's derivation of

divinity fudge and his definition of Syllabication are thus in conflict"

(Lee and Howard 1974, page 221).

(b) Stampe's formulation of Syllabication does not seemingly capture

his original intent. Again quoting here directly from Lee and Howard:

"A significant fact about Stampe's derivation (1) [Figure Three -- PCB]

is that the gradual contraction of the form is the result of deletions

which take place within the stressed syllable. The deletion of [t],

for example, comes about only after other processes have operated to

bring it into final position in the stressed syllable. We gather from

Stampe's discussion that his statement of Syllabication does not cor-

rectly capture his intent, inasmuch as he insists that It] is syllable-

initial and his argument rests heavily upon that assumption" (page 221).


(c) It might be assumed Stampe is claiming that a process of

Flapping in one example involving /t/ results not from an obligatory

syllable-final Flapping process but rather from an optional syllable-

initial process which is only tangentially acknowledged.13 It is just

this assumption that is made by Lee and Howard and for them, therefore,

"certainly the crucial role of the syllable-structure assumption that

is being made merits a clearer statement of this process than we are

given" and "it is difficult to see the justification for syllable-

initial flapping here, since the environment in the doctor (before a

stressed vowel) is quite dissimilar to the one that is relevant in

divinity" (page 222).

(d) Stampe's proposed derivation turns out unnecessarily lengthy.

Lee and Howard have countered that it is sufficiently problematical

that there are additional casual speech processes suggesting a further

syllabication of the nonsyllabics between stressed vowels or diphthongs

within the preceding syllable. A consequence, therefore, of assuming

that /t/ is optionally attached to the preceding syllable is that

processes of syllable-final Flapping and optional Flap-deletion result

in the forms which Stampe has marked unacceptable. A reanalysis of

this data would therefore seem to be in order, and one suggested

reanalysis involving motivation of an obligatory process of Vowel

Syncope yields derivations of the pronunciation of divinity fudge that

are considerably shortened, such as those in Figure Four.























*da.vIn.at. i

d 1.vIr.ar.i

da. v I. i




1. *do.vln.at.i

2. a*d .vlr.ar.i

3. da.vir.r. i

4. da.vI .r.i

6. *da.vIar.i

7. da.vlar.i

8. da.vIIr.i

9. da.vlr.i

where r = English flap

I = front unrounded high vowel

= unpronounceable form (all relevant processes not applied)











If the derivations given in Figure Four are reasonable (a question

on which we might elect to suspend final judgment at least at present)

then Stampe's instances of reapplications of processes in this example

are considerably reduced. But nonlinear applications have obviously

not been strictly ruled out, since two processes of Syllabication and

two of Vowel-nasalization will still apply. Yet Lee and Howard also

have reason for doubting that even these instances provide any sound

argument in support of the case for nonlinearity as Stampe maintains.

One argument stems from the distinct types of English Vowel-nasalization

involved in these separate applications, while another views Syllabica-

tion as a universal rather than a language-specific condition on rules.

In the case of Vowel-nasalization, for example, one application is

progressive while one is regressive. This means "for these to constitute

a valid argument against linear ordering, we would have to claim that

these form a single unitary generalization" (page 224). Lee and Howard

continue further in this vein: "While it could be argued that we are

offering no evidence against such an assumption, there is good reason

to be suspicious of mirror-image environments constituting unitary

generalizations. Aside from Lightner's admonition (1971:234) that the

mirror-image convention predicts linguistic changes of unattested types,

evidence may be given from both synchronic grammars and children's

acquisition to show that such cases often (if not always) have quite

different properties" (page 224).

If the two occurrences of nasalization in Stampe's derivation might

be examples of quite distinct processes once more closely examined, then

processes of Syllabication, after the fashion by which Lee and Howard

assess them, might well also represent distinguishable types. Since

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs