Do speech errors feed phonology?

Earlier today, Geoff Pullum wrote a short post on Language Log about speech errors, based on the following example:

you’re a kind man [knd mæn]
you’re a canned mine [kænd mn]

Geoff concludes his post as follows:

The details of such errors have often been used by phonologists as evidence for phonological structure. After all, if you can accidentally switch the nuclei of two adjacent syllables when you’re very tired, one obvious explanation would be that phonology is not a kind of fiction made up in the process of doing linguistic analysis; rather, there really are syllables, they really have nuclei, and your speech production mechanisms actually operate in a way that, in effect, makes reference to these units.

This made me think about how little I actually know about speech errors and the evidence they provide for phonological structure, apart from the tidbits I discuss in introductory courses (the source of most if not all of these probably being Vicki Fromkin‘s work). In particular, I’m unaware of any examples that might show interaction with phonological rules as opposed to phonological structure.

Now before you write to point out to me that the [k] in the result of the speech error above is more fronted than it is in the source (due to the following front vowel), or that the swapped vowels in the result differ somewhat in length from their respective correspondents in the source (due to the following cluster [nd] vs. single consonant [n]), let me clarify that I’m specifically thinking of rules that cannot be reasonably attributed to the phonetics (coarticulation and so forth) — not that this isn’t in itself interesting, but as this example shows, we already know that speech errors can (and do) feed such phoneticky rules.

I’m more curious about the following types of (completely hypothetical) examples. Suppose you mean to say I worry about Jean’s sanity i:nz sænɪɾi:], and you swap the indicated vowels/nuclei. Will it come out as ænz si:nɪɾi:] (swap only) or as nz sɛnɪɾi:] (swap feeds trisyllabic shortening and vowel shift)?

Or, if that type of example is too questionable, let’s take a language in which high vowels undergo backness and rounding harmony and non-high vowels only undergo backness harmony (like Turkish). Now suppose you mean to say a two-word phrase like [bolumun tukalɯr] (completely made up), and you swap the indicated vowels/nuclei. Will it come out as [bolamun tukulɯr] (swap only) or as [bolamɯn tukulur] (swap feeds vowel harmony)?

My gut-level intuition in these two cases is that English trisyllabic shortening and vowel shift are either too “deep” or non-phonological (in the relevant sense) to be fed by the speech error (so I’d expect the swap-only result ænz si:nɪɾi:]), and that Turkish vowel harmony is “surface-y” enough (again, in the relevant sense, so I’d expect the swap-feeds-harmony result [bolamɯn tukulur]. It’d be interesting to concoct a case that fits somewhere between these two — “deeper” than vowel harmony but more “surface-y” than trisyllabic shortening vowel shift — to see what happens.

Anyone know of any relevant cases, or want to concoct some hypothetical example(s)?

6 thoughts on “Do speech errors feed phonology?

  1. hh

    Hi Eric — Don’t know about constructing Turkish examples, but I was introduced to the wild world of speech errors (admittedly with a morphological, rather than phonological, focus) by Roland Pfau’s extremely interesting 2000 thesis from Utrecht, “Features and categories in language production” Highly recommended reading (plus it makes Distributed Morphology look good, which I always like, too. :))

  2. Eric Bakovic

    Thanks, Heidi — Pfau’s publications webpage is here, and I see that he is preparing his dissertation for publication. (A brief search does not reveal it to be available electronically, not even at the lingBuzz home of the Distributed Morphology archive.)

  3. Karthik Durvasula

    There is some interesting work on speech errors by Goldrick and Blumstein who try to induce speech errors thru made-up tongue twisters. Another paper by Goldrick and Rapp looks at syllabic/non-syllabic information thru aphasic speech.

  4. Marc

    This is a really intruiging way at getting at the productivity of rules.
    One that I can think of off the top of my head would be that metathesis slips in English most certainly feed aspiration allophony – it’s hard to imagine they don’t.
    I can’t think of any attested ones (since I also never learned much about errors) but assuming ‘pEsgeti’ is an error, then I’m sure the ‘p’ gets aspriated. That might not be the best example of metathesis because of the vowels moving around but perhaps there are others…

  5. Eric Bakovic

    Thanks for the comment, Marc. Again, not that this is uninteresting, but I would also classify aspiration with /k/-fronting as “more phoneticky”. I guess you could say that I’m specifically interested in whether speech errors feed non-allophonic rules.

Leave a Reply

Your email address will not be published.