Language Log

« previous post | next post »

As Eric Baković recently noted, there's been a lot of buzz about a presentation about "uptalk" by Amanda Ritchart and Amalia Arvaniti at the 2013 Acoustical Society meeting. All we have so far is a sort of press release  ("Do We All Speak Like Valley Girls? Uptalk in Southern Californian English", ASA Lay Language Papers, 12/5/2013), but this is enough to see that Ritchart and Arvaniti have made a valuable contribution.

They based their analysis on systematic analysis of a good-sized recorded dataset (23 "native speakers of SoCal English", who were asked to describe a muted video clip and to participate in a "map task" interaction). They distinguished among different interactional functions ("simple statement", "question", "floor holding", "confirmation request"), they systematically noted aspects of the location and extent of rises, and they based their conclusions on a statistical analysis of the interrelationship of these features.

Their main conclusions:

For the when of "uptalk", we found:

• Rises were used significantly more often by female than male speakers
• Speakers used rises significantly more often during the map task than when retelling the clip of the sitcom scene
• Rises were always used to ask a question or request confirmation; 45% of floor holding and 16% of simple statements in the sample also ended in uptalk.

For the realization of "uptalk", we found:

• When the rise was used for a simple statement, it began significantly later on in the utterance than when the rise was used for asking a question. In other words, different realizations of the rises were used for different discourse purposes.
• The extent of the rise was significantly greater in questions and confirmation requests than in floor holding; in turn floor holding rises were significantly greater than those used for simple statements. Here again we see that different realizations of the rises were used for different discourse purposes.
• Female speakers began their rises significantly later on in the utterance than male speakers.
• Females used significantly larger rises than males.

From the discussion available so far, it's hard to evaluate these conclusions. We don't know what their criteria were for dividing transcripts into "statements" or other units, or for characterizing units as "simple statements", "floor holding", "confirmation request", etc. We don't know how they defined the starting point and the extent of the rises. And we don't know what they mean by "significantly", i.e. how big the differences really were. This will all be clarified by a fuller presentation of their results, and especially by the (I hope forthcoming) publication of their annotated dataset.

But we do know that it's wrong to imply that this phenomenon is a special characteristic of "SoCal English". The "Map Task" design comes originally from a project at Glasgow and Edinburgh in the 1980s, whose annotated recordings were published in 1993 as the  HCRC Map Task corpus, documented here. And we don't have to listen to more than a few seconds of the first of the 128 sessions to hear final rises of various sorts:

Of course, this might be because the Glasgow speakers in this dataset have the "Urban North British Rise", as well as because the "map task" interaction tends to make speakers more conscious of their step-by-step interaction with an interlocutor. So to get rid of the geographical effect, we can turn to a replication of the map task done in Canada a few years later, and published in 1996 as the DCIEM Sleep Deprivation Study (documented here). Again, the first few seconds of the first of 216 published interactions exhibits various sorts of final rises (the speaker in this case is a 23-year-old male soldier, recorded in 1994 in Toronto):

It would be interesting to compare the distribution of rises in the HCRC and DCIEM map task datasets with Ritchart & Arvaniti's dataset — this would give us a sense of whether there's anything particular to SoCal in this phenomenon.  It's possible (and even likely) that there are differences in the phonetics of the rises, or in their interactional functions, or in their relative frequencies. But without such a comparison, we don't actually have any evidence — beyond stereotypes and anecdotes — of a special connection between these patterns and southern Californians, present or past.

And it's not just "map task" interactions where North Americans of all ages, regions, genders, and eras are likely to use rising intonations. Earlier today ("Once so ever") I noted a rise-replete recorded statement by Elytte Barbour, recently arrested for allegedly joining his bride in a "thrill killing":

Mr. Barbour is young (22) but is apparently from North Carolina, not California.

For a sample of older speakers from an earlier time, I took a random dive into the Santa Barbara Corpus of Spoken American English. The first thing I listened to was the "Bank Products" segment (sbc0014), recorded 12/22/1993. The speaker is "Joe", a 45-year-old white male from Dupo, IL.  He's a loan officer in a small-town bank, and the setting is a loan meeting with two board members. About a third of his phrases end in rises, for example:

From previous experience, I expected that talk-radio-host monologues would be a rich source of rises, and the first Michael Medved clip I sampled (full audio here) did not disappoint:

Michael Medved, born 10/3/1948 in Philadelphia, was 64 years old at the time of this recording.

For an older female speaker, also not from southern California, we can turn to Hillary Clinton's recent speech (video here) accepting the Lantos Prize, in which she eulogized Nelson Mandela:

These last rises may a rather different function from those we've seen so far — although the paired alternatives are connected with "and", and both members of each pair are present with rises, the pairing might be analogous to the "it might be A, or it might be B" disjunctive structures that are often prevented with paired rises and falls. I often find it difficult to assign specific utterances to a classification of rhetorical or interactional function, in part because I'm often tempted to invent a new category, as in this case . So a careful study of inter-annotator agreement (as was done for the "discourse act" classification applied to the HCRC map task corpus) would be a good idea.

For a systematic study of form and function in the (extensive) use of final rises among college-age women Texas in the 1980s, we can turn to Cynthia McLemore's work, described e.g. in "The interpretation of L*H in English", TLF 1991. The implications of this work remain largely unexplored. And here's a 1994 Connie Chung segment that features Dr. McLemore, and uses examples collected by interviewing random (male and female) students at Villanova (near Philadelphia):

I'm always glad to see linguistic research getting extensive media coverage — but the content of the media response to Ritchart & Arvaniti has been more than a little disappointing, for reasons that I hope will be obvious:

Some previous LLOG discussion of related issues:

"This is, like, such total crap?", 5/15/2005
"Uptalk uptick?", 12/15/2005
"Angry rises", 2/11/2006
"The Affect: Sociolinguistic speculation at the NYO", 3/22/2006
"Further thoughts on 'the affect'", 3/22/2006
"Uptalk is not HRT", 3/28/2006
"Satirical cartoon uptalk is not HRT either", 11/14/2006
"Uptalk anxiety", 9/7/2008
"The phonetics of uptalk", 9/13/2008
"Word (in)constancy", 9/16/2008
"Uptalk v. UNBI again", 11/23/2008
"Elementary-school uptalk", 11/30/2008
"Annals of uptalk: the python wrestler", 3/6/2010
"Uptalk awakening", 9/29/2013
"Okie uptalk", 11/10/2013

December 8, 2013 @ 12:28 pm · Filed by Mark Liberman under Prosody

Permalink