# The linguistics behind aave-compress
This tool rewrites prompts using five documented grammatical features of African American Vernacular English (AAVE) โ also called Black English or, in earlier and now-contested usage, Ebonics. Every rule below is traceable to a peer-reviewed source. None of it is invented, and none of it is "slang." AAVE is a systematic, rule-governed variety of English with its own grammar, in the same sense that any other dialect (Appalachian English, Scots, Singlish) has its own grammar โ a claim settled by linguistics, not opinion.
## Why this is a compression mechanism at all
Several AAVE tense-aspect markers pack a distinction that Standard American English (SAE) can only express with a multi-word hedge phrase into a single preverbal particle. John Rickford makes this point directly in "Suite for Ebony and Phonics," citing Toni Morrison's observation that Black English effectively has "five present tenses" where SAE's single "is/am/are" flattens the same distinctions into one form. That is the mechanism this tool exploits: not "fewer words" as an end in itself, but grammatical features that were already doing more semantic work per morpheme.
## The five markers used in v1
### Zero copula (copula deletion)
**Rule:** Wherever SAE can contract "is"/"are" (*"he's tired"*), AAVE can delete it entirely (*"he tired"*). Where SAE cannot contract ("I don't know where he **is**"), AAVE cannot delete either โ this is the empirical test that proves the deletion is rule-governed, not random dropping.
**Source:** Labov, William. "Contraction, Deletion, and Inherent Variability of the English Copula." *Language* 45(4): 715โ762, 1969.
### Habitual/invariant "be"
**Rule:** Uninflected "be" marks habitual or iterative aspect and cannot be negated with "ain't," inverted for questions, or used in a tag question โ it behaves unlike any SAE auxiliary. *"He working"* (right now) is grammatically distinct from *"He be working"* (habitually/regularly) โ a distinction SAE needs an adverb to make.
**Source:** Labov 1969; Fasold, Ralph. *Tense Marking in Black English*. Center for Applied Linguistics, 1972.
### Completive/perfective "done"
**Rule:** Preverbal "done" + past-tense verb marks that an action is completely finished, typically with added emphasis or a note of surprise that SAE's neutral perfect ("have already finished") doesn't carry as compactly. Unlike the SAE perfect, AAVE "done" can co-occur with definite past-time adverbials like "yesterday."
**Source:** Green, Lisa J. *African American English: A Linguistic Introduction*. Cambridge University Press, 2002; Spears, Arthur K. "The Black English Semi-Auxiliary Come." *Language* 58(4), 1982.
### Prospective "finna" (fixing to)
**Rule:** A prospective/immediate-future modal marking that an action is imminent, functioning as a distinct, ethnolinguistically-marked form โ not simply a contraction of "going to."
**Source:** Green 2002; Thomas, Erik R. and Grinsell, Timothy. "Finna as a Socially Meaningful Modal in African American English." Presented at SULA 7, Cornell University.
### Continuative-intensive "steady"
**Rule:** A preverbal adverb marking an activity as persistent, intense, and consistent โ continuative aspect layered with an intensifying (sometimes indignant) flavor. Baugh's original example: *"Ricky Bell be steady steppin' in them number nines"* โ persistently and intensely stepping out.
**Source:** Baugh, John. *Black Street Speech: Its History, Structure, and Survival*. University of Texas Press, 1983.
## What's documented but deliberately NOT used for compression
- **Remote-past stressed "BIN"** โ real compression value (a single stressed word disambiguates "started long ago and still true" from ambiguous SAE "has been"; Rickford's 1975 comprehension test found 92% of Black respondents vs. 32% of white respondents correctly read "She BIN married" as still-married), but the entire distinction rides on a capitalization signal that a downstream logger, proxy, or auto-formatter can silently destroy. Held for a v1.1 "advanced" opt-in once we've verified a target model's behavior actually changes on the signal rather than pattern-matching the lexical item regardless of case. Source: Rickford, John R. "Carrying the New Wave into Syntax: The Case of Black English BIN," 1975.
- **Negative concord** and **existential "it's"/"they"** โ both real, rule-governed, and documented (Labov, "Negative Attraction and Negative Concord in English Grammar," *Language* 48, 1972; Green 2002) but token-neutral or marginal in isolation. Not worth the misparse risk for the savings they'd add.
- **Signifyin' / semantic indirection โ excluded on purpose, not an oversight.** Claudia Mitchell-Kernan's foundational account ("Signifying and Marking: Two Afro-American Speech Acts," 1972) and Henry Louis Gates Jr.'s literary elaboration (*The Signifying Monkey*, Oxford University Press, 1988) describe a "double-voiced" mode where a single utterance is built to carry two simultaneous meanings โ one literal, one ironic or critical โ and correct interpretation depends on the listener doing real inferential work. That is the opposite of what a reliable, low-ambiguity instruction prompt needs from an LLM. It's real, it's one of the richest features of Black discourse, and it has no place in a compression engine whose entire job is to reduce ambiguity, not require it.
## AAVE is not "broken English" โ the record on this is closed
On December 18, 1996, the Oakland Unified School District Board of Education passed a resolution recognizing Ebonics as a distinct, rule-governed linguistic system, triggering national controversy built substantially on a misreading of what the resolution actually proposed. In response, the Linguistic Society of America passed a resolution โ drafted by John Rickford and adopted unanimously at its January 1997 business meeting โ stating plainly that AAVE "is systematic and rule-governed like all natural speech varieties," and that characterizing it as "slang," "mutant," "lazy," "defective," "ungrammatical," or "broken English" is "incorrect and demeaning." That resolution is this project's north star for tone: the pitch here is that AAVE makes *more* grammatical distinctions than SAE in the places this tool touches, not fewer.
**Source:** Linguistic Society of America, "Resolution on the Oakland 'Ebonics' Issue," January 3, 1997 (full text: https://web.stanford.edu/~rickford/ebonics/LSAResolution.html); Smitherman, Geneva. *Talkin and Testifyin: The Language of Black America*. Wayne State University Press, 1977.
## Full bibliography
- Baugh, John. *Black Street Speech: Its History, Structure, and Survival*. University of Texas Press, 1983.
- Fasold, Ralph. *Tense Marking in Black English*. Center for Applied Linguistics, 1972.
- Gates, Henry Louis, Jr. *The Signifying Monkey: A Theory of African-American Literary Criticism*. Oxford University Press, 1988.
- Green, Lisa J. *African American English: A Linguistic Introduction*. Cambridge University Press, 2002.
- Labov, William. "Contraction, Deletion, and Inherent Variability of the English Copula." *Language* 45(4), 1969.
- Labov, William. "Negative Attraction and Negative Concord in English Grammar." *Language* 48, 1972.
- Linguistic Society of America. "Resolution on the Oakland 'Ebonics' Issue." 1997.
- Mitchell-Kernan, Claudia. "Signifying and Marking: Two Afro-American Speech Acts." In *Directions in Sociolinguistics*, ed. Gumperz & Hymes. Holt, Rinehart & Winston, 1972.
- Rickford, John R. "Carrying the New Wave into Syntax: The Case of Black English BIN." In *Analyzing Variation in Language*, 1975.
- Rickford, John R. and Rickford, Russell John. *Spoken Soul: The Story of Black English*. John Wiley & Sons, 2000.
- Smitherman, Geneva. *Talkin and Testifyin: The Language of Black America*. Wayne State University Press, 1977.
- Spears, Arthur K. "The Black English Semi-Auxiliary Come." *Language* 58(4), 1982.
- Thomas, Erik R. and Grinsell, Timothy. "Finna as a Socially Meaningful Modal in African American English." SULA 7, Cornell University.
- Yale Grammatical Diversity Project โ phenomena entries for "Invariant be," "Perfective done," "Steady," "Fixin' to," "Stressed BIN (been)," "Negative concord," "Expletive they." https://ygdp.yale.edu/