Language Log

In a comment on "Reading Instruction in the mid 19th century", Rachel Churchill asked

Does contrastive emphasis (as in "George or his brother") exist in all languages? If not, which ones don't have it?

I've sometimes noticed non-native English speakers – even those whose pronunciation and accent are pretty good – failing to use it. For example, they might say "This one is fifty GRAMS, but the other one is twenty GRAMS", where a native speaker would emphasise the "fifty" and "twenty" rather than the "grams". I'm guessing it's because their native language doesn't use contrastive emphasis and maybe they've never been taught the concept, but I don't know this for sure.

I responded:

See Yong-cheol Lee, Bei Wang, Sisi Chen, Martine Adda-Decker, Angélique Amelot, Satoshi Nambu, and Mark Liberman, "A crosslinguistic study of prosodic focus", IEEE ICASSP, 2015. The abstract:

We examined the production and perception of (contrastive) prosodic focus, using a paradigm based on digit strings, in which the same material and discourse contexts can be used in different languages. We found a striking difference between languages like English and Mandarin Chinese, where prosodic focus is clearly marked in production and accurately recognized in perception, and languages like Korean, where prosodic focus is neither clearly marked in production nor accurately recognized in perception. We also present comparable production data for Suzhou Wu, Japanese, and French.

See also "Victor Hugo, hélas", 4/13/2024, "LÀ encore…", 4/14/2024, and "Intonational focus", 4/21/2011.

Roger C. continued the discussion:

In Spanish, focus contrast is generally achieved by changes in word order; syntax is more flexible than in English.

And I responded:

I'm guessing that (as in French) the word-order changes are combined with tonal and durational changes, and that in some cases (as in corrective focus on number or letter strings), the prosodic changes are the only cues.

Commenters noted the ambiguity of this sentence quoted earlier today in "Rococo":

When President Donald Trump returned to the White House in January, he wasted little time redecorating.

From Bob Ladd: "I was genuinely uncertain when I read the sentence about 'wasting little time' whether Trump had in fact gone right to work redecorating or rather had decided not to bother.

There are several different types of "fronting" or "preposing" in English, sometimes categorized in syntactic terms (e.g. wh-movement) and sometimes in pragmatic terms (e.g. topicalization). Here's recent example of a familiar type, for which I don't know a standard name:

The stage was set for Tesla to get its launch, and its launch it got.

That example seems a bit awkward to me, but definitely still possible. Examples where the preposed item is a simpler noun phrase seem to go down a bit easier — for example, substituting "a launch" for "its launch".

"Does GPT-4 Surpass Human Performance in Linguistic Pragmatics?" Bojic, Ljubiša et al. Humanities and Social Sciences Communications 12, no. 1 (June 10, 2025). Ljubiša Bojić, Predrag Kovačević, & Milan Čabarkapa.  Humanities and Social Sciences Communications volume 12, Article number: 794 (2025) Cite this article

Abstract

As Large Language Models (LLMs) become increasingly integrated into everyday life as general-purpose multimodal AI systems, their capabilities to simulate human understanding are under examination. This study investigates LLMs’ ability to interpret linguistic pragmatics, which involves context and implied meanings. Using Grice’s communication principles, we evaluated both LLMs (GPT-2, GPT-3, GPT-3.5, GPT-4, and Bard) and human subjects (N = 147) on dialogue-based tasks. Human participants included 71 primarily Serbian students and 76 native English speakers from the United States. Findings revealed that LLMs, particularly GPT-4, outperformed humans. GPT-4 achieved the highest score of 4.80, surpassing the best human score of 4.55. Other LLMs performed well: GPT-3.5 scored 4.10, Bard 3.75, and GPT-3 3.25; GPT-2 had the lowest score of 1.05. The average LLM score was 3.39, exceeding the human cohorts’ averages of 2.80 (Serbian students) and 2.34 (U.S. participants). In the ranking of all 155 subjects (including LLMs and humans), GPT-4 secured the top position, while the best human ranked second. These results highlight significant progress in LLMs’ ability to simulate understanding of linguistic pragmatics. Future studies should confirm these findings with more dialogue-based tasks and diverse participants. This research has important implications for advancing general-purpose AI models in various communication-centered tasks, including potential application in humanoid robots in the future.

…and not just for women:

I've based several past posts on passages from Benjamin Franklin's Autobiography. Recently I happened to notice some large differences among different editions, and so I took a look at the Library of Congress page "Finding Benjamin Franklin: A Resource Guide", which lists 16 "significant" early editions of Franklin's Autobiography, as well as a guide to that work's "complex early history". From that I learned that, well, its history is complicated — which is also clear from the Wikipedia entry. But in the course of making some textual comparisons, I happened on a passage that (in all its variants) raises the question, what did Franklin have against Edinburgh?

Kase Wickman, "Kylie Kelce Knocked Joe Rogan Off the Top Podcast Slot, In Just One Episode", Vanity Fair 12/10/2024:

Kylie Kelce, wife of Jason Kelce, sister-in-law of Travis Kelce, mother of three (with a fourth on the way), and Dunkin Donuts enthusiast, can add another descriptor to that incomplete list: No. 1 podcast host.

The debut of Kylie’s new podcast, Not Gonna Lie, has unseated Joe Rogan’s The Joe Rogan Experience from its long-held perch atop the most-listened charts on both Apple and Spotify. The inaugural episode, featuring It’s Always Sunny In Philadelphia star Kaitlin Olson as a guest, was published on December 5, with new episodes planned to drop each Thursday.

A recent Elle Cordova short:

Ian Prasad Philbrick and Ashley Wu, "The 9 Elements of a Trump Rally", NYT 10/8/2024:

The energy for Mr. Trump’s third White House campaign comes from his rallies. Since President Biden dropped out of the race and Vice President Kamala Harris took the helm, Mr. Trump has held nearly 20 of them, speaking for about 90 minutes at each.

Like most politicians, he repeats things at every speech. Unlike most politicians, he offers a grim view of the country, makes up nicknames for his opponents and pledges to use the power of the government to punish his rivals.

To help readers experience what a Trump rally is like, we used video to break down the nine themes he consistently returns to.

Screenshot of a post on Threads:

Marilyn Singer is reponsible for the local (re-)invention of "Reverso Poetry: Writing Verse in Reverse":

A reverso is a poem with two halves. In a reverso, the second half reverses the lines from the first half, with changes only in punctuation and capitalization — and it has to say something completely different from the first half (otherwise it becomes what one blogger’s kid called a “same-o.”)

Wikipedia uses the term "Reversible poem", and tells us that

A reversible poem, also called a palindrome poem or a reverso poem, is a poem that can be read both forwards and backwards, with a different meaning in each direction, like this:

Reversible poems, called hui-wen shih poems, were a Classical Chinese artform. The most famous poet using this style was the 4th-century poet Su Hui, who wrote an untitled poem now called "Star Gauge" (Chinese: 璇璣圖; pinyin: xuán jī tú).This poem contains 841 characters in a square grid that can be read backwards, forwards, and diagonally, with new and sometimes contradictory meanings in each direction.[2] Reversible poems in Chinese may depend not only on the words themselves, but also on the tone to produce a sense of poetry. Beginning in the 1920s, punctuation (which is uncommon in Chinese) was sometimes added to clarify Chinese palindromic poems.

The focus of this post is Brian Bilston's Reverso Poem "Refugees".

Noor Al-Sibai, "AI Systems Are Learning to Lie and Deceive, Scientists Find", Futurism 6/7/2024:

AI models are, apparently, getting better at lying on purpose.

Two recent studies — one published this week in the journal PNAS and the other last month in the journal Patterns — reveal some jarring findings about large language models (LLMs) and their ability to lie to or deceive human observers on purpose.

Matt Taibbi has gotten some teasing for mixing metaphors in a recent Xeet about Bannon's jailing:

That’s . . . a lot of metaphors.

[image or embed]

— Radley Balko (@radleybalko.bsky.social) Jun 7, 2024 at 6:42 PM