口音里的身份

越不想跟学校站一边的孩子，元音变得越快

八十年代初，语言学家 Penelope Eckert 在底特律郊区一所她化名为 Belten 的高中泡了两年。学生自己分成两拨：jocks，围着学校转，进学生会、打校队、打算上大学；burnouts，心思在校外、在底特律城里，多来自工人家庭。她录下他们说话，一个元音一个元音地量。

当时美国北方城市里正在扩散一串元音变化，叫北方城市元音推移（Northern Cities Shift）——比如 but 里那个元音往口腔后部滑，听起来越来越像 bought。结果 burnouts 在这串变化上普遍走得比 jocks 远；走得最远的是 burnouts 里最铁杆的那批女生。在 I didn't do nothing 这类双重否定上，她们甚至超过了所有男生。

这些差别没有一个是「没学好」：同一所学校、同一批老师、同样从小说英语。差别出在想往哪边站。往学校那边站的，口音贴着课本那条线；往城里那边站的，口音跟着城里走。元音成了一件穿在身上的队服。

口音还能在几秒钟里被调大。1977 年，Bourhis 和 Giles 在威尔士找了一批正在学威尔士语的成年人，让一个说英格兰上层口音的人在语言实验室里提问，其中一句挑衅：你们学一门「前途暗淡的垂死语言」干什么？听到这句，受试者的威尔士口音明显加重，有人往回答里直接塞威尔士语单词。这叫分化（divergence）：故意把自己说得跟对方不一样。反方向的趋同（convergence）更常见——跟谁聊就慢慢往谁的说法上靠，开头那通电话就是。

所以语码转换（code-switching）除了语法那一面（「两种语言住在一个脑子里」那期讲过换点的规矩），还有社会那一面：换说法就是在换「我此刻跟谁站一起」。证据强度：Eckert 是一所学校的深度田野调查，Bourhis 和 Giles 是一次小样本实验，两者都被当经典反复引用，但单个研究的规模都不大。

EN In the early 1980s Penelope Eckert spent two years in a Detroit-suburb high school she called Belten. The students sorted themselves into jocks, oriented to the school, and burnouts, oriented to the city and mostly from working-class homes. On the Northern Cities Shift — a chain of vowel changes then spreading through northern US cities, such as the vowel of but sliding back toward bought — burnouts were generally further along than jocks, and the most committed burnout girls furthest of all; on double negatives like I didn't do nothing they outdid every boy. None of this is failure to learn: same school, same teachers, all native English speakers. The difference is which side you want to stand on. Accents can also be turned up in seconds: in Bourhis and Giles's 1977 study, Welsh learners challenged by an upper-class English voice about studying "a dying language with a dismal future" broadened their Welsh accents and some dropped Welsh words into their answers — divergence, the opposite of the more common convergence toward whoever you are talking to. So code-switching has a social face as well as a grammatical one: changing how you talk changes whom you are standing with right now. Evidence strength: one in-depth school ethnography and one small experiment, both much-cited classics, neither large.

一个 hello，就够把你分进一栏

最早的记录在《圣经·士师记》里：基列人守住约旦河渡口，让过河的人说 shibboleth。以法莲人的方言里没有开头那个音（像英语 she 开头那个音），说成了 s（像英语 see 开头那个音），一开口就露了底，当场被杀。这个词后来在英语里就指「一开口就暴露身份的那个暗号」。

1999 年，语言学家 John Baugh 和同事把这件事做成了实验。Baugh 本人能说三种美国英语：标准美式英语、非裔美国人英语（美国黑人社群的一种英语，有自己成体系的语法）、奇卡诺英语（墨西哥裔美国人社群的英语）。他用这三种说法给旧金山湾区几个城镇登租房广告的房东打电话，每次说同一句话，问报上登的那套房子。

另一组实验把录音剪得只剩开头那个 hello，让听众猜说话人属于哪个族群——大约七成猜对。机制就摆在这里：口音能从极短的一段声音里被可靠地认出来，而它又跟族群、出身、阶层绑在一起，于是「听口音」成了在不准明说的地方做分拣的一条路。Baugh 给它起名叫语言画像（linguistic profiling）。看不见脸，并不等于公平。

更制度化的版本在难民审查里：一些国家请人听申请者的口音，判断他来自哪里。2004 年，十九位语言学家联名发布一份指南，核心是一句话——口音能告诉你一个人在哪个说话社群里长大，告诉不了他拿的是哪国护照。索马里语就同时在索马里、吉布提、埃塞俄比亚、肯尼亚说，国界从说话社群中间穿过去。

EN In the Book of Judges, Gileadites guarding the Jordan fords made travellers say shibboleth; Ephraimites, whose dialect lacked the opening sh sound, said s, and were killed on the spot. In 1999 John Baugh and colleagues turned this into an experiment. Baugh, who can speak Standard American, African American and Chicano English, phoned San Francisco Bay Area landlords in each voice with the same enquiry about an advertised flat. In Woodside the standard voice got a viewing about 70% of the time, the other two about 29% and 22%; in Palo Alto 63%, 48% and 32%. Cut down to the single word hello, listeners still identified the speaker's group about 70% of the time. Accent is identifiable from a tiny slice of sound and is tied to ethnicity, origin and class, so listening to accent becomes a way to sort people where sorting them openly is forbidden — what Baugh named linguistic profiling. The institutional version is accent analysis in asylum cases; the 2004 guidelines signed by nineteen linguists say accent can tell you which speech community someone grew up in, not which passport they hold — Somali is spoken across Somalia, Djibouti, Ethiopia and Kenya.

听不懂的那一半，在听的人那边

2019 年，Taylor Jones 等人请来费城法院的 27 位书记员——在庭上逐字记录证词的人，约占全市书记员的三分之一——让他们转写一批非裔美国人英语的日常句子。这份工作的执照要求准确率不低于 95%。实际上，整句完全写对的平均只有 59.5%，最好的一位 77%，最差的 18%；不少人事后也说不出那句话是什么意思。

错的不是说话人含糊。非裔美国人英语有标准英语没有的语法区分。语言学家 John Rickford 七十年代做过一个经典测试：问「她结婚了吗」，答 She BEEN married，BEEN 重读。多数黑人受访者听成「早就结了，现在也还是」，多数白人受访者听成「结过，现在不是了」。同一句话，两套语法读出相反的事实；放进证词里，它就是判决依据。

听力是能练的。Bradlow 和 Bent 2008 年让以英语为母语的人先听一段带汉语口音的英语，再测他们能不能听懂一个从没听过、同样带汉语口音的人。只跟一个说话人练过的，换个人优势就没了；跟好几个说话人练过的，优势能带到新的人身上。「听不懂」很大一块，是听的人没听过足够多样的这种口音——一次训练就能让它移动。

可现实里，被要求调整的几乎总是说话那一方：去报班改口音、面试时换成标准说法。语码转换的社会面又回来了，只是这回是被迫的。「谁的话算标准」那期讲的是评分怎么偏；这里是更早的一步——理解的力气，默认该由谁来出。

EN In 2019 Taylor Jones and colleagues had 27 Philadelphia court reporters — about a third of the city's pool, certified at 95% accuracy — transcribe everyday sentences in African American English. Only 59.5% of sentences were transcribed fully correctly on average, best 77%, worst 18%, and many reporters could not say what a sentence meant. The speakers were not unclear; the dialect has grammatical distinctions standard English lacks. In John Rickford's classic test, asked whether she is married, the answer She BEEN married with stressed BEEN was heard by most Black respondents as "married for a long time, and still is" and by most white respondents as "was, but no longer" — opposite facts from one sentence, which in testimony become grounds for a verdict. Listening can be trained: in Bradlow and Bent's 2008 study, English listeners trained on one Chinese-accented talker gained nothing with a new talker, but those trained on several carried the benefit to a new one. Much of "can't understand" is a listener who has not heard enough variety, and one session moves it. Yet it is almost always the speaker who is asked to adjust — the social face of code-switching again, now compelled. The topic on whose speech counts as standard was about biased rating; this is a step earlier: who is expected to do the work of understanding.