A discussion of the vagaries of prestige subject pronoun usage can be framed by comparing it to a consistent alternative usage. As linguists widely acknowledge, English-speaking children all seem to acquire the usage of the examples — before certain of them are exposed to corrections which reverse or partially reverse the prescriptively in­correct judgments in —.

It may well be that middle class negative concord can be successfully internalized by the speaker of working class American with only a moderate amount of training, given motivation on the part of the learner. In fact, the grammatical difference between working class and middle class negative concord, in spite of the strong difference in "flavor" between the examples of , is absolutely minimal and easy to grasp, once a precise formal framework such as generative grammar is available for succinct statements of the appropriate rules.

It strikes me that the demystifying implications of the early generative analyses of phenomena such as negative concord were in fact a source of panic for the mentally fogged purveyors of proper English, and caused them to rush for shelter under the imprecise and supposedly socially aware rubrics of sociolinguistics, language in context, etc. Under these "approaches", facts, anecdotes, and correlations can be gathered for­ever without fear that the flimsiness of language "barriers" between classes and groups will be actually understood. Today, the educational and linguistic establishment have made these "humanistic" approaches to language the order of the day, and the idea that generative analyses can show how minimal and unimportant most socially-based language differences are has been buried: cf. the program for the 1985 Summer Institute of the Linguistic Society of America.

In what follows, the separate constructional types exemplified in - will be discussed in more detail. Here, it suffices to note that the unstarred non-prestige patterns in - are the quite general unschooled adult usage, besides being a stage through which "schooled" children who eventually acquire some "correct usage" pass. For purely expository reasons, let us refer to the usage of - as "normal" (NU).

An informal description of uncorrected American NU of subject pronouns can be derived from Klima (1964b). The subject pronouns I, we, he, she, and they are used as a noun phrase (NP) if and only if the phrase is an immediate constituent of a sentence (S) which contains an inflected verbal element. Let us assume that the subject pronouns in English are represented in a speaker's lexicon with the features PRONOUN, SUBJECT, while the other (non-possessive) forms are listed simply as PRONOUN. Further, let INFLECTION be the category that is realized alternatively as the modals or the finite present and past tense endings of English. We can propose the following formalism for NU: NU rule: PRONOUN - INFLECTION → [ PRONOUN ] [ SUBJECT ]
- INFLECTION Using more formal transformational notation and the simplification of note 3, (11a) can be alternatively stated as (11b): NU rule: PRONOUN – 1 → [1,2] - 2

Rule formalizes most of , but we must still guarantee that applies only to pronouns that are immediate constituents of S. In much recent generative work, the INFLECTION category is argued to be the "head" of S. The head of a phrase is that grammatical category in a phrase which is obligatory and which selects or "governs" its

INFLECTION or "I" is then the category that inverts in English questions, appears in tag questions (e.g., John is late, isn't he? John can swim, can't he?), precedes n't and emphatic so/too, and remains when the verb phrase is deleted or understood (e.g., Mary couldn't leave town before Bill I[did]; Whether or not Bill I[can], you should visit New York.)

Rules that need not concern us here insure that the verbs be, have, and do are realized in the I position in surface structures of sentences (cf. Emonds, 1976, Ch. 6).

The ad hoc feature SUBJECT in rule below can be identified with I, so that becomes a "copying rule," with no loss in generality. In another work (Emonds, 1985, Ch. 3), I is further identified with the category SPECIFIER(V).

Since an element cannot govern outside the smallest maximal phrase which contains it, PRONOUN cannot govern INFLECTION. (Personal pronouns do not tolerate sister constituents within the same NP: *the they, *that younger she, *it in New York, etc; arguments and explanations of apparent counterexamples are given in Emonds (1976, 118-119).) Therefore, in order for to apply, INFLECTION must govern PRONOUN. Even though the pronouns in are adjacent to INFLECTION, they are not governed by the latter, and so by universal grammar , they cannot be realized as subject pronouns. The stories about (us/*we) are good. None of (them/*they) can help. Travis's condition therefore guarantees that rule will turn only pronouns which are immediate constituents of S into subject pronouns.

It might seem surprising to many advocates of prestige usage that normal usage, which violates "correct grammar," is fully systematic and rule-governed; moreover, this usage is of interest to theoretical linguistics, in that it suggests how to formulate principles of universal grammar so as to both exclude the non-occurring variants in and also allow the normal usage of -.

Let us now examine how , as a local and language-specific transformational rule conforming to universal grammar, generates the normal usage in -.

Conjoined Subjects. Chomsky's (1973) principle of subjacency (or alternatively, Ross's (1967) coordinate structure constraint) prevents The head of a phrase is the central member of the phrase because it is obligatory (in the usual cases), because its category determines the category of the smallest phrase that contains it, and because the head is "selectionally dominant". This latter term means that when the phrase is subject to co-occurrence restrictions with material outside the phrase, it is the properties of the head which most centrally figure in these restrictions. Thus, in the castle's six-foot six gatekeeper that you spotted, the noun gatekeeper (or keeper) is obligatory, determines that the phrase is a noun phrase, and enters into co-occurrence restrictions; for example, a verb which requires an animate object such as frighten can occur with an object whose head is gatekeeper. We say that a head B governs another head C if the phrase of which C is the head is a complement of B; in terms of trees, if the C-phrase is a sister of B. from affecting the conjoined Noun Phrase ("NP") subjects exemplified in . The circled bounds for subjacency in constitute the "double bound" that prevents constituents from being related by a transformational rule, so pronouns in conjoined subjects cannot be subjective in form.

Subjects of Understood Predicates. In , even though the pronouns are understood as "subjects," they are not followed by an overt INFLECTION which governs them, and hence Travis's condition that one term of a local transformation govern the other blocks . The pronouns therefore appear in their unmarked or "elsewhere" objective form (me, us, him, her, them). We will see below that in the appropriate abstract syntactic structures for the sentences in , the pronouns are probably in fact paired with "empty" INFLECTION nodes as in . However, the lower INFLECTION (I) in , although it is in a position to govern the NP her, does not dominate a phonologically realizable syntactic morpheme. That is, the lower I is empty, and transformational rules typically cannot "see" (more formally, "analyze") categories which are not realized in the terminal syntactic string. Thus, the I in is invisible to , and the higher INFLECTION does not govern the pronominal NP, so in neither case can apply to yield subject pronouns in NU.

Predicate Nominals. In , the explanation of why her and us appear in NU follows from the fact that explicitly requires the subject pronouns which are induced by a governing INFLECTION to be to the left of I, and yet predicate attributes are to the right of I.

First Person Demonstratives. Rule for NU does not require absolute surface adjacency between inflection and PRONOUN, as the possibility of certain intervening adverbs and parentheticals in NU shows: We (*Us) usually are the ones to blame. Harry thinks that she (*her), rich as she is, should pay for this. However, as stressed above, does require that INFLECTION govern PRONOUN, and thus that the pronoun be the head of the subject NP (cf. note 4). In the examples of , the first plural pronominal form we/us is not in fact the head of the construction, but rather substitutes for a demonstrative determiner (these/those); as such, it is not governed by INFLECTION, but by the head noun of the subject NP if by anything. In the first example of , the head governed by INFLECTION is commuters (cf. 17a), and in the second it is empty (cf. 17b). These commuters are often blamed for smog, Those (ones) with insurance have to pay a lot. Therefore, since the pronouns in do not satisfy the Adjacent Head Condition, they remain in objective form in NU.

Appositives to Subjects. It is not clear whether appositive NP's are generated as parentheticals and fall under the case of isolated subjects with understood predicates, or whether they are sisters to the NP they modify in an NP[NP-NP] structure, as argued in Delorme and Dougherty (1972). In the latter case, appositive NP's are structurally like the second conjunct in a coordinate NP. In either case, the pronoun in an appositive NP is not an "adjacent head" to INFLECTION, as required by , so does not apply, and NU will show object pronouns, as in .

In principle, NU should allow us to decide between these alternative analyses of appositive NP's. If Delorme and Dougherty are correct, an NP subject modified by an appositive NP, being analogous to a conjoined NP, should exhibit object pronouns in NU, whereas if appositive NP's are parenthetical, the modified NP should not be affected and should, even in NU, exhibit subject pronouns. Even though the test should be clear-cut, a problem arises because the use of appositive noun phrases, especially after pronouns, may well itself be associated with a prestige "formal" or business style, and thus induce unwitting sociological conformity to prestige usage (i.e., a change of "register") even among speakers of NU. (Prestige usage of course requires subject pronouns, whichever analysis of parentheticals is correct.) Relevant but inconclusive data for which NU usage should be determined are as in ; John said that (she / her), his best friend, could never get the job. Now (we / us), the renters in this district, are going to have to pay more.

To terminate this section, I conclude that the normal usage NU of subject pronouns in all five of the constructions just discussed, as observed in American working class speech and in the untutored speech of American middle class children, is fully systematic and in accord with universal grammar. NU accords with such principles as subjacency, the adjacent head condition, and the characteristics of local transformations. In particular, the subject pronouns of NU are generated by the minimal and elegant language-particular transformational rule . It remains to be seen in what way prestige subject pronoun usage can be described and/or justified in terms of an equally consistent grammatical framework.

Although the handbooks (HCH and CW) vary in details, a rough idea of what is considered English prestige dialect subject pronoun usage (termed here PU) can be obtained by reversing the grammaticality values in -.

Prestige usage is typically defended on the grounds that the grammatical notion of "subject" determines when a subject pronoun should be used. For example, from HCH: "choose the case form that shows the function of pronouns or nouns in sentences" and "use the subjective case for the complement of the verb be" . By these criteria, if the notion "subject" can be correctly understood, the prestige usage of - follows.

In fact, subject pronoun PU can quite plausibly be related to well-founded concepts in the theory of grammar. (This is not to say that it conforms to them, as will be seen below.) In recent grammatical research, a theory of "abstract case" has been developed, by means of which all phonologically realized NP's (= "lexical NP's") must be assigned one of a small number of mutually exclusive case features which in large part, but not always, correspond to the grammatical relations that these NP stand in with governing heads such as V (verb) and P (preposition). In languages which exhibit relatively complete systems of morphological case endings (e.g., Classical Greek, Finnish, German, Latin, Modern Standard Arabic, Russian), it is to be expected that abstract and morphological cases will largely if not wholely coincide. Moreover, it is claimed that abstract case on all lexical NP's occurs in all languages, by universal grammar, and hence even in a language where nouns are not inflected for case, such as English.

Among the abstract cases is "nominative case," which is taken to be assigned to lexical subject NP's by the head of S, namely INFLECTION. In my own work on abstract case (Emonds, 1985, Ch. 5), I propose that the nominative case feature can be identified with the category INFLECTION itself, just as note 3 identifies the features SUBJECT and INFLECTION. That is, the universal theory of abstract case has the invariable effect that a lexical subject NP is assigned the "case feature" I ( = INFLECTION).

Work on abstract case generally agrees that case is assigned at the same level of representation to which the local transformation applies. (In Chomsky (1981), this level is "s-structure.") However, there are crucial differences between the way the general abstract case-marking of universal grammar assigns I to a subject NP and the way that the language-particular local rule assigns I: Abstract case-marking applies only to phrases (e.g., NP). Without some further statement, abstract case is not even realized on classes of morphemes. Rule , on the other hand, applies directly to the morpheme class PRONOUN. ` Abstract case-marking "percolates down" to all conjoined immediate constituents of a phrase it is present on. ` Since abstract case-marking is not a transformation, but an independent principle of grammar, it is not subject to subjacency (the coordinate structure constraint) like is. Abstract case-marking applies to the subject NP of a syntactically empty INFLECTION with an understood predicate. Thus, the subjects of the understood predicates in and are abstractly nominative. In contrast, a local transformation can only analyze (see) an INFLECTION which is not empty. The characteristic of "linking verbs" like be is precisely that they cannot assign abstract "accusative case" (= the category V, in my view) to their complements. When a linking verb appears, the theory of abstract case guarantees that an NP or AP complement is assigned the next highest available case category (i.e., nominative). Thus, universal abstract case is not purely directional in its effect, as is a local transformation like with a left-to-right structural description. Languages often contain rules which specify certain "agreements" between a head noun and the determiner that modifies it. Thus, in many languages in which abstract case is realized morphologically on N, it is also realized on the corresponding DETERMINER. A local transformation involving I such as can, however, by Travis's condition , only affect the head of the subject NP, and not a modifier. Abstract case-marking must universally assign, at least optionally, the case of an NPi to any other NP in apposition to NPi. For a study of optional morphological case agreement on appositive NP's in German, see van Riemsdijk (1983).

The above characteristics of a universal theory of abstract case are discussed in more detail in Emonds (1985, esp. sections 1.8, 5.7, and 5.8). The differences between abstract case assignment of nominative and assignment of nominative by a local transformation, essentially all theorems in a properly formulated universal grammar, are what explain the discrepancies observable in - between NU and PU. When we realize that the ancestors of Modern English, Old English and Early Middle English, were morphologically case-marking languages, in which abstract case was realized similar to the way it is in Modern German, it becomes clear what prestige usage is. PU is simply the attempt to claim that the case on pronouns in Modern English corresponds to that of Old English, i.e., that a universal theory of abstract grammatical case, to be validated on the basis of its ability to describe German, Latin, Russian, etc., also determines the distribution of subject pronouns in English.

In fact, translation into German of the examples in - shows that the case distinctions of everyday spoken German mirror those of the prestige dialect of English. In contrast to English, however, the choice of subject and object pronouns poses no difficulty for the native speaker of German.

The question thus becomes, is it or is it not a possible and indeed plausible principle of grammar, that a universally valid theory of abstract case can determine the distribution of subject pronouns in English, just as it does in languages with extensive systems of morphological case?

An intellectually honest appraisal of PU in English speakers must at least conclude that pronoun usage based on abstract case is not easily acquired. Let us say, intuitively, that the reason for this is that far fewer forms in English (only pronouns) distinguish between subjects and objects. In languages like German, morphological case distinctions appear also on nouns and on the determiners and adjectives that modify nouns. (E.g., German determiners agree with nouns not only in number, as in English, but also in case.) That is, the properties of the abstract case system are transparent in German morphology, because they are realized in so many different forms. This difference between German and English case can be expressed formally as follows. Definition. A syntactic category C is "morphologically transparent" on B if and only if a productive number of pairs of simple B which contrast with respect to C also differ phonologically.

According to , using the definitions in note 6, an example of a category which is transparent on both English N's and English NP's is ± PLURAL. If one is willing to say that the English singular mass nouns permit a phonologically empty indefinite determiner that cor­responds to the article a(n), then the existence of productively many pairs like bread / a bread, gas / a gas, and fluoride / a fluoride indicate that the feature ± COUNT is syntactically transparent on English NP's but not on English N's. Finally, the feature ± ANIMATE is transparent neither on English N's nor on English NP's.

Consider now the category of subject pronouns in English, where the C in is the feature SUBJECT (i.e., INFLECTION) and B is NP. Assuming PU, the rules of English yield only five pairs of simple NP's which differ both phonologically and by virtue of being ± SUBJECT (namely, I/me, etc.). Thus, the feature SUBJECT is not morphologically transparent on English NP's. In contrast, German contains an abundance of simple NP's which contrast syntactically only in case features and still contrast phonologically; for example, any noun phrases which contain a masculine singular head noun and a modifying determiner or adjective will differ phonologically according to whether they are abstractly nominative (they contain SUBJECT = INFLECTION) or accusative. That is, the category "nominative" is morphologically transparent on German NP's ( der Apfel/den Apfel 'the apple, ein Junge/einen Jungen 'a boy,' jeder Student/jeden Studenten 'each student,' etc.).

We can translate the earlier intuitive description of the learning difficulty of English PU into a more formal property by using morphological transparency: Morphological Transparency. An abstract (e.g. case) feature C of a category B is realized on the lexical head of B in a language if and only if the C is morphologically transparent on B. With respect to our topic here, means that the subjective case feature INFLECTION cannot be realized on the head of English NP, since INFLECTION is not morphologically transparent on NP.

If the evidence to be examined below actually suggested that PU were a dialect of natural language, a weaker statement than would be in order; for example, one could claim that the realization of a case feature which is not morphologically transparent is simply difficult to learn, rather than unlearnable. However, the next section establishes sociolinguistically that PU is not internalized by any speakers of English, and is the principle of Universal Grammar which explains why this is so. moreover explains why local transformations like come into existence. If a generation of speakers for whom abstract case has become morphologically opaque is to maintain the semblance of the pronoun system of a linguistic "older generation," they must invent a rule like , which in turn, as a transformation, necessarily fails to accurately reproduce the patterns induced by abstract case. If the proper distribution of a morphologically opaque category were simply difficult to learn, sufficient correction could lead to a construction of an internalized grammar expressing this category, and one might witness even the re-invention of PU by later generations. However, there is not the slightest hint that linguistic change is reversible in this way, especially with regard to English PU for subject pronouns. Given this reality, the attempt at maintaining PU by prescriptive grammar and the educational system leads only to a linguistic tool for maintaining instead socio-economic class barriers. Those with access to advanced education and paid secretarial services can effortlessly intimidate, confuse, and disorient those without such access.

Before turning to sociolinguistic considerations in the next section, I will mention some further evidence that is essentially correct as a general principle of universal grammar. For example, since the category P is closed (= not productive) in at least all the languages I have seen discussed, there cannot be a productive number of simple PP's in a language. By definition , no category can then be morphologically transparent on PP in any such language. If some abstract feature analogous to a case feature is assigned to a PP from outside the PP by some rule of grammar, it then follows from that this case(-like) feature cannot be realized inside PP. And indeed, while there are case-like particles that can be positioned at the periphery of PP's in various languages like Japanese and Persian (e.g., the particle ez of Persian studied in Samiian, 1983), no language with morphological case realized on heads and/or determiners (e.g., the Indo-European languages, Modern Standard Arabic, Finnish) exhibit a similar "case" inside PP's on P itself.

A second example of is furnished by the agreement of a verbal past participle in French with a preceding cliticized, relativized, or questioned direct object pronoun. Les tartes que j'ai cuites pour la soirée. 'The pies (fem.) which (fem.) I have baked (fem.) for the party.' Quelle journée! II me l'a décrite de façon drôle. 'What a day (fem.)! He to me it (fem.) has described comically. This rule transfers the marked ("feminine") grammatical gender of a phonologically null direct object "trace", which is to the right of the verbal past participle, to that participle (a VERB). Since only a small class of irregular past participle stems are phonetically marked by this rule (-(s)crit, -peint, -vert, -pris, -cuit, and about ten others) Morphological Transparency can be applied as follows: The abstract gender feature FEMININE on V is realized on the lexical head (i.e., the same V) if and only if there is a productive number of V which contrast phonologically according to whether they are grammatically feminine; since the latter is not the case in Modern French, the rule of past participle agreement cannot be maintained if the participle is a V.

As could be expected from , then, the rule of past participle agreement is not internalized by French native speakers; the Académie Française admitted to this reality in the late seventies after its centuries-long battle to save this rule.

In contrast, a rule which is "alive" in French syntax assigns the feminine gender to these same past participles when they follow a feminine subject and the passive auxiliary être 'be'. Past participle agreement in the French passive manages to reproduce itself in internalized grammars because it is part of a more general rule that assigns feminine gender to a productive class of adjectives (after être and other verbs). (It is morphologically evident in many Romance languages that the past participle in the passive construction is an adjective, while in the "composed past" forms, it is a verb.)

The contrast in French between which agreement rule survives and which one dies out thus further supports the correctness of Morphological Transparency as a criterion for learnability.

I will now present the evidence that PU, even though claimed by many American native speakers to be their own, is not learnable. Needless to say, if only the (obviously conscious) acceptability judgments of prestige dialect speakers are consulted, PU will be judged a dialect, since they will often insist that, in their family or social circle, PU really is the natural way to speak.

Three fundamental sociolinguistic factors give the lie to this pretense and indicate that PU is not simply a dialect difference between middle and working class speech (and thus akin, for example, to different negative concord usage). They are: ) College and business writing handbooks, of the type consulted by secretaries, technical and ghost writers, journalists, copy-editors, etc., typically include large sections on subject pronouns, but dismiss negative concord as presenting no difficulties. ) These handbooks, as well as casual observation of "educated speakers," testify to widespread "overcorrection" in pronoun usage. The handbooks zealously combat this overcorrection, but at the same time treat imagined overcorrection for negative concord as a disease of schoolteachers not worthy of (at least masculine) attention (cf. below).

The crucial role of Morphological Transparency in eliminating Object-past participle agreement in French is testified to by the existence of such a rule in Catalan, where a productive class of past participles are phonetically differentiated by the rule.

The French rule exhibits an interesting adjacency (locality) requirement between the participle marked and the trace involved. For example, in les tartes quei j'ai fait cuire ti, pour la soirée 'the pies whichi I have had bakedi for the party', the participle fait must not agree; nor can it agree if the trace is from an indirect rather than a direct object.

As far as I can see, this rule (in its living versions) is a strong argument for the trace theory of movement rules, since the agreement is indifferent to the position of the antecedent of the trace but not to that of the trace.

These facts follow directly from my hypothesis that PU is not a phenomenon of internalized natural language. My reasoning is thus: I first establish below that the handbooks adequately represent PU and that they are destined precisely for adult (i.e.. 18 or over) native speakers of the prestige dialects of American English. If these speakers have prestige negative concord but not PU of subject pronouns in their internalized grammar, , and then follow. follows from my hypothesis and the further fact that whatever prestige subject pronoun usage is acquired (i.e.. through means other than normal language learning) cannot be generalized by using abstract case features to a full range of subject and predicate attribute NP's; hence, in many linguistic contexts any attempt to use subject pronouns in place of the object pronouns simply leads to the production of novel ungrammatical utterances which are unacceptable to the native ear. Since approval of object pronoun usage in these essentially formally unspecifiable contexts would demonstrate the futility of rationalizing the whole prestige sham, the handbooks must recommend "avoidance."

Linguists out of contact with the teaching of English might think that the handbooks are a dusty reminiscence of a past age and of past attitudes, and no longer meet some demand of the market. But this would Just before my three-year-old internalized middle class usage of any, he went through a brief period when I got any shoes meant I got no shoes, showing that "overcorrection” of double negatives is quite otherwise from what CW imagines. be the wishful thinking of the ivory tower. The Harbrace College Handbook has gone through eight editions in twenty years, and the author of The Careful Writer was the style editor for The New York Times for about twenty years. These handbooks further present themselves as accurate reflections of present usage, though it is of interest to note that they (as opposed to many middle class people) make no claim that anybody speaks in the manner they prescribe: HCH sets out to "describe the usage of good contemporary writers" (p. viii—my italics), and CW is intended "as a guide to good written English usage" (p. vii—my italics).

The contemporary nature of the handbooks is also shown by the basic agreement between HCH and CW, with respect both to content and to what is stressed, and by the fact that they claim to contain "the principles of effective writing. These include (as has been shown by a comprehensive examination of student writing) everything to which instructors normally refer in marking papers" (HCH, p. viii).

It is not only that the books present themselves (and supposedly are bought) as codifications of present and necessary "educated usage." The educational system presents such books to the American population in the same way. These two books were, within the last ten years, the only two on English grammar at the reserve desk at the University of California at Berkeley library. Anyone who comes to this highly reputed library with a general question on English grammar is likely to be guided to the general reference room and the books discussed here. Without doubt, the standards set by handbooks such as these are part of today's American social and educational reality.

One might object that these handbooks are in fact intended for speakers of non-prestige dialects, and hence are meant at least as aids to some kind of upward social mobility: this objection would then suggest that prestige subject pronoun usage, the native usage of some segment of the middle or upper class, is a realistic target for self-improvement. But the explicit statements and the contents selected in these books show that these books are not aimed at teaching the speech of prestige dialect. For the handbooks do not in general give rules or make assertions about any other linguistic usage that is clearly internalized only by prestige dialect speakers.

For example, the handbooks completely avoid discussing two characteristics of prestige American which are different from working class speech. Working class negative concord referred to earlier is simply taken by CW as "gutter language" and a sign that "such people should be in another business" (presumably, where "careful writing" is not required, e.g. labor). Rather than discussing the rules for middle class usage, CW launches into an offensive attack on (women) teachers who supposedly overcorrect "double negatives." HCH devotes three lines to the double negative, except that HCH also points out that certain lexical items such as scarcely are lexically negative. But this lexical correction can be "useful" only under the assumption that for the corrected speaker a lexical negative will automatically not co-occur with not. This unstated assumption is thus equivalent to assuming that HCH's intended reader has middle-class negative concord to start with, exactly my point.

Another example: American working class speech often exhibits an undifferentiated past and perfect participle form. The handbooks ignore this, consistently addressing themselves to middle class speakers. One can contrast this gap in the handbooks, destined for prestige dialect college students and graduates, with a "pedagogical" ridicule aimed at working class students lacking two distinct past forms during my education in American high and grade schools.

If the handbooks are not meant for speakers of non-prestige American dialects, they are a fortiori useless for non-native speakers of English. Thus, they say nothing about when to use an infinitive and when to use a gerund in English—presumably a difficult task for the non-native. (CW simply counsels "parallelism" in coordinate structures in such cases, begging the question entirely.)

The handbooks are destined then for American prestige dialect speakers. Moreover, a glance at their style and at their market (journalists for CW, college students or graduates for HCH) convincingly shows they are for adults. Finally, these books are not only "glossaries" of misunderstood words, even though in fact much of the information they contain is purely lexical and is appropriately arranged in lists of "misused words." Nor are they summaries of the principles of effective writing, even though they again contain some material of this sort. Rather, the first and emphasized parts of these books are devoted to grammar, a grammar that the authors consider practical and necessary in everyday writing.

The handbooks can therefore be taken as current, and as answering some need of the marketplace; furthermore, this need is one of adult speakers of the prestige dialect. That is, there is something grammatical that prestige dialect speakers don't know how to do, which the handbooks claim to help them with.

It is not that these speakers are ignorant of the difference between subject and object, or that they confuse morphological forms, for errors are not produced in simple sentences like . What they don't know is how to reproduce PU consistently in the constructions listed in -—conjoined subject NP's, subjects of understood predicates, etc. The handbooks are most persistent in requiring subject pronouns in conjoined subjects (e.g., the PU forms in , which are the starred NU forms). However, they spare no effort in combating NU in other constructions. For example, they are very concerned with PU in predicate nominals. The two handbooks being cited dedicate complete sections to this topic, giving both simple examples like and predicate nominals in cleft sentences, as in . There will be only we two at dinner. (CW, 352) I'm sure it was he who did it. (CW) Another favorite target of correction consists of subjects of understood predicates in examples like , where PU is insisted on. John did that yesterday, not I. (CW, 352) No one will show you more of the West Indies and South America in 7 days than I. (example from a Costa Line advertisement) Everyone knows the answer except I. (CW treats at length whether except is to be followed by a subject pronoun.) Finally, the handbooks are quite ready to correct examples such as the following, which are taken from sources which deliberately mimic the NU of nonprestige speakers. Us hookers get a lot of flack, (pulp fiction) Us Tareyton smokers would rather light than fight, (ad copy, caricaturing "hardhat" or "cowboy" speech)11

If the handbooks' only grammatical task seemed to be correction of certain NU, that of object pronouns being "overused," one might claim that they are aimed at, say, healing a dialect split among otherwise linguistically uniform PU speakers. Their assigned task in the market would then just be a calling to order of that segment of prestige speakers who are slipping into NU with respect to only one difference of internalized grammar (subject pronouns), while another segment produces full PU with no problem. However, the handbooks' treatment of subject and object pronouns is not limited to correcting NU, as this usage is reproduced in -. An equal amount of rhetoric is devoted to combatting an overextension of subject pronouns into object positions, which is not a characteristic of NU. An earlier study of NU (Klima, 1964a) has noted: "The usage in L4 (NU, J.E.) agrees entirely with that of the dialogue in Nelson Algren's The Man with the Golden Arm and The Neon Wilderness." This is to say, authors who reproduce children's and working class people's "rough" or completely untutored speech do not replace both subject and object pronouns of PU with the "wrong" forms; rather, only certain subject pronouns are so replaced, while all object pronouns of PU are also object pronouns in NU.

Therefore, the combativeness of the handbooks concerning the replacement of PU object pronouns with the wrong subject forms (e.g., they prefer not to think about James and I more than necessary) is directed not at pure NU speakers nor at NU. It is directed rather at PU speakers or (even more importantly) at would-be PU speakers who are making innovations in the English pronoun system which are neither PU nor NU. These innovations the handbooks strongly stigmatize with the term "overcorrection," and it is to this topic that we now turn.

When speakers of a language internalize a construction in that language, they do not, after the period of acquisition itself, persistently overgeneralize the construction into paradigms where it is not acceptable. For example, English-speaking adults do not overgeneralize the passive as in (36a), number agreement as in (36b), quantifier "float" as in (36c), or auxiliary inversion as in (36d). A month's salary was cost by my coat. The boy ates dinner. My friends have three gone home. Got John examined by a doctor? Nonetheless, English speakers who exhibit some aspects of PU in spontaneous speech invariably extend the use of subject pronouns to NP's which do not have the abstract nominative case that accurate prestige usage requires. Prescriptive grammarians and the handbooks call this phenomenon "overcorrection." Some typical examples of overcorrection (i.e., extension of subject pronouns to non-nominative NP's) are given in . They prefer not to think about James and I more than necessary. She and Bill we ought not to mention. John told no one other than I. Who, if not he, could we have hired? It was she that John discussed in detail. Mary expected the messenger to be she who had called earlier. We democrats the voters expect to be more enlightened. For they to be understood correctly, . . . (produced by a College English professor). As a speaker trained in PU, I find that all of (37a-g) sound "refined" to my ear, and yet all are completely "incorrect."

Perhaps the type of overcorrection most discussed among "educated speakers" is that involving use of subject pronouns in coordinate objects of verbs and prepositions (cf. 37a-b). It is generally acknowledged that this error is frequent, and quite decisively incorrect—a sure sign that PU is being violated. The speakers who err in this way never, in contrast, produce subject pronouns in simple object phrases (*they prefer to think about I; *she Sue ought not to mention; etc.).

Responding to this common version of overcorrection, HCH devotes its first two sections under "case" to coordinate subject pronouns, attacking overcorrection in object and other positions. CW takes subject pronouns in coordinate NP's as typical of "overrefinement," citing examples such as between you and I and let's you and I go to the theatre.

How can we explain overcorrection, especially in the light of the non-occurrence of errors as in ? If I am correct in postulating , partial acquisition of PU can be internalized only by adding more local transformations, formally similar to , to the grammar. These transformations are ad hoc attempts to mimic certain PU paradigms, which are themselves ungrammatical; moreover, these local rules are "too general", in that they immediately give rise to hosts of examples as in which can be stigmatized by the prestige dialect reader if they should be produced in writing. A sample of what such rules might be is as follows: A pronominal NP following and, or, but, than, as is subjective (cf. , c). A pronominal NP modified by a relative clause is subjective (cf. ). A pronominal NP following a finite form of be is subjective (cf. ). A pronominal NP in sentence-initial position is subjective (cf. , g,h). A pronominal NP which is an appositive is subjective (cf. , g). It can be noted that in accord with Morphological Transparency , all the above statements are local transformations and in no way utilize the abstract case features on an NP; that is, they are formally like the rule which characterizes NU of subject pronouns.

No prescriptive grammarians countenance the rules in , and this is exactly the rub. Many ordinary English speakers deprived of higher education manage to internalize some set of rules similar to those in , and to produce some kind of mixture of NU and PU. Now, since the use of forms generated by rules as in is exactly what they unconsciously realize sets them apart from NU, they naturally enough can conclude that their speech and writing is prestigious (i.e., they are not "talking like kids"). And when this speech is heard or read by the judging members of the dominant socio-economic class, they are caught℄recognized for what they are; they can be singled out as "unlettered" as the need may arise, precisely for having produced what they felt was "correct."

If prescriptive grammarians actually wished to unveil PU for what it is, and to make it accessible to lower class speakers, they would explain it as an unlearnable system, and one which can be consciously understood only in terms of abstract case. They would then point out that rules like those in are precisely what upper class speakers use in unguarded speech, and exactly how these rules can be pitfalls. But can the prescriptive grammarian admit that PU usage depends on a hodge-podge of inelegant and unprincipled makeshift strategems to protect a device used to reinforce class differences, and that PU has nothing to do with logic, grammar, semantics, or intelligence? It would seem not, because their audience would no doubt ask: why should the PU of subject pronouns, or for that matter prescriptive grammar, even exist, since it is not part of natural language?

One might imagine a defense of PU which claims that English speakers maintained the pronoun system based on abstract case for centuries after Morphological Transparency predicts that this system should have been lost. However, this defense is quickly shattered by the realization that overcorrection in examples like those in has occurred throughout the Modern English period; most of the examples in are recorded in Fowler's treatise on prestige usage (which incidentally has the merit of not taking itself as seriously as the handbooks). All debts are cleared between you and I. (Shakespeare) Wagers lost and won between him and I. (Pepys) Twixt this good man and I. (Bunyon, Pilgrim's Progress, 308) Whether ... is not for we outside mortels to decide. (Fowler, 1965, 689) To discover only one solitary person, and he a sentry, on the steps . . . (Fowler, 1965, 78) I saw a young girl . . . whom I guessed to be she whom I had come to meet. (Fowler, 1965, 78) Let us be content—we Liberals, at any rate—to go on . . . (Fow­ler, 1965, 669)

I have been able to confirm the overcorrection and general inability to master PU of subject pronouns in a questionnaire distributed to 65 students in introductory linguistics courses at the University of Washington. These students are for the most part prestige dialect speakers—exactly the audience that the handbooks are destined for, both during their college writing careers and later during middle and upper-middle class careers. Their average age was 20.5 years, and of course only native English speakers were tested.

The subjects were given five minutes to mark 25 sentences as grammatically correct or not grammatically correct. 12 of the sentences didn't involve controversial pronoun usage, while 13 utilized pronouns in contexts where various PU errors were at least conceivable. The 12 "non-pronominal" sentences included instances of negative concord and the past/past participle distinction, such as the contrasts in -. Which drugs have you took lately? Which bottle was shaken the hardest? Who come in yesterday? Mary scarcely wrote any articles. Nobody done very well. Nobody took the course. Didn't they bring no books back in? Didn't they say no books came in? We didn't see no one there. Hasn't anyone eaten yet? Of the 780 (12 x 65) grammatically judgments given on the 12 non-pronominal sentences, 82.4% (544) accorded with prestige dialect speech, even though, as pointed out earlier, the constructions used in these sentences are far from uniform in the American speech community. Moreover, on the 5 non-pronominal examples which violated prestige usage (that is, the starred examples in -), the percentage of judgments that accorded with prestige usage was 90.5% (294 of 325); this indicates that the test-takers were extremely well aware of what violates PU in the non-pronominal constructions.

When the 845 judgments given for the 13 examples involving potentially controversial pronoun use are examined, it turns out that only 60.9% (515) agreed with prescriptive grammar, a number that is barely higher than chance. 8 of the examples on the test, reproduced as , involved overcorrection (i.e., one type of violation of PU), and only 51% (265 of 520) of the responses indicated an awareness of these violations; that is, these college students overcorrect as much as they do not. When the contexts for overcorrection are present, namely, those where the ad hoc local transformations of violate prescriptive grammar, these prestige dialect speakers are completely incapable of more than chance behavior in attempting to reproduce PU. Who, if not he, could we have hired? We discovered only a single person, and he a guard, on the steps. They prefer not to talk about James and I more than necessary. It was she that John discussed in detail. I expected the young girl to be she whom I had seen before. Who do you like, if not they? Let us be content, we liberals at any rate, to continue our fight. John told no one other than she.

The results of this test then support my claim that adult native speakers of the prestige dialect of English have not internalized subject pronoun PU. Not a single test-taker reproduced prescriptive pronoun usage without error, while 10 made no errors on the non-pronominal test sentences; 59 of 65 made more than 2 errors on pronoun usage, while only 32 made more than 2 errors on the non-pronominal usage. The rampant overcorrection of these speakers on just one type of rule is inexplicable if we assume that PU can be acquired as an internalized grammar, but is easily understood as an interplay between some or all of the patterns induced by and an awareness, especially in a test situation where written language is involved, that conscious "grammatical principles", as partially inculcated through a middle-class upbringing, are to be brought into play. Thus, the internalized grammars of prestige dialect speakers, no less than that of NU speakers, assign "case" to English pronouns only by means of local transformations. They cannot utilize abstract nominative case, which in English is a morphologically opaque phrasal feature incapable of realization on English morpheme classes such as PRONOUN. The only differences between the internalized grammars of PU and NU speakers is that the former add on ad hoc local transformations as in to supplement the local transformation which is common to all dialects of American English.

The handbooks, as well as any native speaker somewhat familiar with PU, are aware of the fact that insistence on PU will, in many syntactic contexts, lead to sentences that are quite unacceptable to the native ear. For example, subject pronouns in predicate nominals can be outlandishly infelicitous: Someone will have to pay for the car, but it won't be likely to be we. Has it been they cooking fish? Is that Mary? It could never be she with glasses. Bill is quite confused; now he's sure that John's I. Othello is a wonderful role; I should be he in your production. I'm never he when we're on tour. She's never seen her cousins; how does she know we aren't they? It could very well be we that they claim left too early. Substitution of object pronouns in for the predicate nominals yields acceptable (NU) sentences, and in fact most prestige dialect speakers also accept the NU variants if they can be presented without calling attention to their grammatical status.

However, the sentences in so clearly offend the native speaker's intuition that HCH proposes an exception to PU, saying that the complement of "non-finite" be can be in the objective case in "informal style." In fact, the exception still only covers half the examples of . As a result, the HCH can offer no definite resolution for these problems. If HCH mean "modal - be" to be finite, so as to cover a greater number of the examples in , they are contradicting CW.

Actually, while wishing to insist on the standard atrophied business responses (this is she, it was I, etc.), HCH is simply relabeling P as "formal usage" in roughly those contexts where PU sounds most unnatural, thereby suggesting that PU may be abandoned essentially arbitrarily (from a grammatical point of view). No rule other than an inherent feel for middle class mores is given to help the reader distinguish "formal" social contexts ("adhere to PU") from the suddenly acceptable "informal" ones ("forget PU").

Both handbooks admit that in some situations PU is stilted or clumsy, and that constructions where only NU and that constructions where only NU sounds acceptable and yet conflicts with PU should be avoided. This might call to mind the way that style manuals frequently suggest avoiding constructions which are unduly long, which interrupt a train of thought, or whose syntax violates semantic parallelism. But these latter considerations do not apply to choices of pronouns. For a writer or speaker within NU (e.g., as is Algren in the books cited earlier), there is no stylistic imperative to avoid the NU counterparts to the examples of , or of others to be discussed below. What is wrong in , and what the handbooks are trying to avoid, is the spontaneous negative grammaticality judgments emanating from the internalized grammars of PU speakers — judgments which conform to NU. Thus, HCH's rule for complements of non-finite be in "informal style" is nothing more than a recognition that the extra local transformation for subject pronouns in predicate attributes () works for many PU speakers only after a finite be. Since their subject pronouns in predicate attributes are not due to abstract nominative case, many of them spontaneously recognize PU after non-finite be as ungrammatical.

In general, syntactic factors never or almost never necessitate "avoiding a construction." The syntax of languages consistently provides a natural way, or several natural ways, for a given message to be expressed. The study of style in syntax is a study of choices, not a study of how to express what is semantically clear but syntactically difficult. Given a semantic "need," syntax is in general adequate to the task.

Therefore, when prescriptive grammarians claim that a thoroughly good syntax is unavailable under certain conditions (i.e., when they counsel avoidance), they are in fact admitting that "bad syntax" (tabooed grammaticality judgments) is lurking just below the surface. The unwanted judgments of native speakers reveal a truth which prescriptive grammar cannot openly admit — that English PU for subject pronouns is not reproduced by an internalized use of abstract case.

Once this truth is out, it must be concluded that PU is reproduced only socially, in a way that is alien to people's spontaneous use of their native tongue. The mode of reproduction is as follows: The prestige speakers internalize a set of ad hoc and artificial local transformations such as those in which mimic, far from perfectly, the distribution of pronouns that would be determined by abstract case theory (e.g., as in German or Early Middle English). In addition, a group of socially appointed experts (prescriptive grammarians, of the "practical sort"— those who control business and journalistic writing, such as is prescribed by the New York Times and Harcourt Brace) devise "grammatical standards" based more or less on how abstract case works, in accord with Morphological Transparency, in languages such as German and Latin. Written English, in order to be socially acceptable, is then supposed to conform to these standards. Should speakers of English without access to the privileges of business (secretaries, ghost writers, etc.) happen to venture into the business world using the unconsciously acquired local transformations that set off their speech from working class usage, they can always be caught (and in writing!), since their use of these transformations will conform to universal grammar, rather than to prescriptive grammar. Hence, they will "overcorrect" the grammatical standard, and be stigmatized.

The only hitch in the social reproduction of PU is this: since the local transformations of PU speakers do not and cannot reproduce the PU standard, and since all native speakers of English have either NU or NU supplemented by these extra local rules, most speakers spontaneously recognize the ungrammaticality of many PU sentences, such as those in , which no version of an internalized English grammar generates. Since the prescriptivists maintain a fiction that they are the authority on speaking grammatically, they must enforce a ban on sentences of PU which are generally recognized as not English; hence, "avoid the construction."

It is not only in predicate attributes that we find examples of PU which prescriptivists must suggest avoiding. In the other constructions which differentiate NU from PU, we can find patterns of PU which sound unacceptable. Conjoined subjects: The Wilson's and we have always been neighbors. Mary thinks that her husband and she spend too much. Are your friends or he going to pick up John? A question for politicians: how necessary are judges or they in a democracy? Sometimes Bill and we were late. Subjects of understood predicates: Students smarter than she get no scholarship. Everyone but they gets on John's nerves. Let's not award a boy as rich as he any scholarship. Bill gave the same gifts as I to her parents. They don't tell people less influential than we the answer. Our friends like robots for presents, but not he. NP's in apposition to subjects: The three New Yorkers, Mary and they, know the most about art. They said that the best cook in town, I, should go shopping. Mary said that the good players, mainly we, had been chosen.

Finally, it is worth observing that the handbooks clearly fail in their "educational" task, that of getting the typical prestige dialect speaker to consistently write and speak on formal occasions in conformity with PU. The continued appearance of overcorrection in formal spoken English and in written English demonstrates that the social code of PU is not uniformly observed. (Cf. the results of the test discussed earlier; also, anecdotally: a playwright on National Public Radio discusses a character's conflict "between she and her daughter.") Indeed, this explains the longevity and the number of editions of books such as HCH and CW. As pointed out above, the very impossibility of successfully imposing PU is actually part of the system which makes PU of subject pronouns such an effective tool for discrimination. The more difficult the mastery of PU, the more PU remains the reserve of those with access to the best secretarial and editing services. An occasional lapse from this arbitrary secret language can be and is tolerated, since what counts is the very difficulty of access. In contrast, where there is a violation of internalized middle-class dialect, something to which the unmoneyed might in principle have access, the condemnation is stronger, immediate, and total (e.g., CW's charge of "gutter language" when middle class negative concord is not observed).

It should be noted that correct prestige pronoun usage cannot be determined by any putative principles of "logic" or "semantics" which are arrived at without reference to the detailed workings of grammatical theory. For that matter, neither can normal usage; as the prescriptive grammarian points out, it is "illogical" to use subject pronouns for the subject in and object pronouns for the subjects in .

An area where all versions of English, including PU and NU, are "illogical" can be observed by comparing two English constructions which, in the contexts where both are possible, are logically and also to a great extent grammatically similar. In one of these constructions, pronominal subjects are subjective in form, and in the other, objective. These two constructions are the present subjunctive and the for-to infinitive. As shown in (i), these two constructions are nearly identical in meaning and would certainly be classed together by any non-grammar-based logical or semantic analysis; moreover, they together contrast in meaning with the indicative clauses shown in (ii). It is absolutely necessary that he (*him) not be late. It is absolutely necessary for him (*he) not to be late. It's important that she (*her) have enough food. It's important for her (*she) to have enough food. Do you prefer that I (*me) be here during interviews? Do you prefer for me (*I) to be here during interviews? As for the chairs, I asked that they (*them) be put away. As for the chairs, I asked for them (*they) to be put away. It's important that she has enough food. Do you prefer (it) that I am here during interviews? As for the chairs, I asked if they were put away.

The arguments that the English present subjunctive and for-to infinitive have the same grammatical status for the most part follow from characterizing both constructions as lacking the expansions of INFLECTION found in indicative clauses (i.e., modals and present/past tense). As a result, neither construction shows any subject-agreement, both use the form be, both reject n't and place not before all verbs, both exclude contraction of have to 've, and both exclude the auxiliary do. Moreover, both the present subjunctive and the for-to infinitive share a meaning of roughly "unrealized modality." The fact that the pronouns in the subjects of these two constructions uniformly differ, while the subject pronoun form appears both in subjunctives and in the semantically contrasting indicatives of , demonstrates that a fundamentally syntactic distinction determines the distribution of subject pronouns. The question is, what is this distinction?

Traditional grammar sheds no light on this question, for it, and also the handbooks, assumes that the objective form of pronoun subjects of infinitives is due to a special addendum to the rules for case (e.g., HCH, 50).

I think we can go beyond an ad hoc stipulation of this sort. In Emonds (1976, section 5.9), four arguments are presented that the subject NP in the for-to construction is attached via a local, language-specific transformational rule to the introductory subordinating conjunction for. This element is structurally outside S, and hence this rule of "for-phrase formation" leaves the subject position inside S "empty" in surface structure, as in : s[[for - NPi] - s[npi [∅] - INFLECTION - VP]] Under this analysis, syntactic terminal PRONOUN elements in the for-to construction are outside S and constitute a surface complement of the head for. Such PRONOUN neither govern nor are governed by INFLECTION, and so rule cannot apply to the sequence PRONOUN — INFLECTION, by Travis's principle . Thus, subject pronouns do not appear, and the expected unmarked variants in objective form serve as subjects.

One may still ask how the pronominal subjects in subjunctives undergo , since transformations only apply to sequences of syntactic terminal elements in trees, and the subjunctive INFLECTION is not phonologically realized. The answer, I believe, is that the absence of INFLECTION in subjunctives results from a deletion of a modal which applies after rule . That is, at the level at which operates, the "s-structure" of recent Chomskyan work, the INFLECTION node is not empty. Indeed, in the British English counterparts to the subjunctives in , the modal should (≠ obligation) is phonologically present. Its deletion in American English is best described as a dialect-specific morphological rule ("should has a zero allomorph in American when obligation is not expressed"). As is typical of most generative models presently in use, such a rule must follow all syntactic rules including . The difference between the syntax of present subjunctives and for-to infinitives is just that induced by the syntactic rule of for-phrase formation which "bleeds" the operation of rule at s-structure. (For additional justifications of this rule, cf. Emonds, 1985. Ch. 7.)

To establish the main point of the appendix, that the form of subject pronouns is not governed by some syntax-independent semantics or logic, it is not necessary to agree on the exact nature of what exempts for-to infinitives from (in both PU and NU). The initial discussion of and , prior to introducing , makes clear that semantics or logic cannot predict the pronominal forms in question.

Another construction in which non-syntactic explanations fail to predict prestige subject pronoun usage is provided by . Everyone but they is being applauded. Prescriptivists typically claim that them in is incorrect, since the pronoun is understood as a subject. The meaning of , however, is identical to that of . Everyone besides them (*they) is being applauded.

Traditional grammar and one handbook (HCH, 193) unhesitatingly classify besides as a preposition taking objective case, even though it is almost indistinguishable in meaning (and logic) from the but in . It therefore appears that, if prestige subject pronoun usage is principled at all, the principles involved are those of grammar and not purely logical or semantic in nature.