A Cupertino of the mind

« previous post | next post »

Yesterday, Stefan Valdimarsson wrote to tell me about an interesting error in one of my recent posts. It was a typing error, but not one of the common slips of the finger that have been catalogued, counted and modeled over the decades, from D.D. Lessenberry's 1928 "Analysis of Errors" (published by Corona Typewriters, and reprinted in Dvorak et al., Typewriting Behavior, 1936) to the "Glossary of Terms Including a Classification of Typing Errors" by D. Gentner et al., in W.E. Cooper's Cognitive Apects of Skilled Typewriting, 1983.

This wasn't a keystroke substitution error, nor a transposition of two sequentially adjacent keystrokes, nor an interchange of keystrokes that are not serially adjacent, nor a migration of keystrokes to a position earlier or later than the canonical order, nor a keystroke omission, nor a keystroke insertion, nor an abstract doubling error (like "aad" for "add") or alternation error (like "threr" for "there"). Such errors are a fascinating subject, as you can learn by reading David Rumelhart and Donald Norman's seminal paper "Simulating a skilled typist: A study of skilled cognitive-motor performance", Cognitive Science 6(1) 1-36, 1982.

But this wasn't really a keystroke error at all.

Instead, it was a case where I started to type one word, and then, as my attention shifted downstream, my fingers continued with a different, and entirely inappropriate, alternative. I was transcribing some experimental instructions from a (.pdf image of) a paper on "Most Comfortable Loudness for Pure Tones, Noise, and Speech" (discussed in the post "Liberman on Sax on Liberman on Sax on hearing", 5/19/2008). The target passage read like this:

This switch enables you to make the speech louder or softer. Using this switch, your task is to make the speech louder or softer until you reach a level which you feel is your most comfortable listening level …

But what I actually typed was (emphasis added):

This switch enables you to make the speech louder or softer. Using this switch, your task is to make the speech louder or software until you reach a level which you feel is your most comfortable listening level …

Stefan's note to me about this struck an indirectly ironic note, under the Subject line "The new address of Language Log?":

To: Mark Liberman
1 Language Log Plaza
CA 95014

I found your recent posts on Dr. Sax's treatment of scientific data very thought provoking. His triumph over the disjointness of the confidence intervals clearly suggests he doesn't understand what he is talking about at all. Thanks for your work in posting this.

But there was this one sentence in your first post which struck me for a completely different reason and made me wonder whether the new address of Language Log might be as above.

Stefan's joke evokes a different sort of error type, for which Ben Zimmer has popularized the term "The Cupertino Effect". (For a list of Language Log Cupertino posts, see here). These are errors where a spell-checking program has guessed wrong about how to revise a typed word that is missing from its lexicon, often transmuting a keystroke error or a foreign import into comedic gold. Thus "the pan-european system of diagonal cooperation" may be rendered as "the pan-european system of diagonal copulation", or "the Muttahida Quami Movement " as "The Muttonhead Quail Movement".

The thing is, though, I type without a net. No (external) spellchecker can be blamed, in this case, because none was involved. I just read (and thought) "louder or softer", but typed "louder or software". (The fact that I didn't detect the error on scanning the post is yet another piece of evidence that I'm the world's worst proofreader.)

Like most other people, I make typing mistakes of this type fairly often. The commonest ones involve mis-continuations of derivation (like "cooperation" instead of "cooperated") or inflection (like "cooperating" instead of "cooperates"). "Software" instead of "softer" could be construed as involving a shared morpheme soft, but I'm not sure how active the association is for me, and this may instead be an example where the target and the output are related only by a shared sequence of initial letters. Such mis-continuations are not nearly as common as keystroke errors are, but I certainly notice them fairly often (or have readers point them out to me, since as I said I'm a lousy proofreader).

In Rumelhart and Norman (1982), these are called "capture errors". They give examples such as "normal" rendered as "norman" and "efficiency" rendered as "efficient".

There's a good deal to be said about such capture errors. The first and most obvious observation is that my substitution of "software" for "softer" must indeed have been a sort of Cupertino, but rather a "Cupertino of the mind", where noise in the transfer from eye to fingers led to a word-level substitution, based on some sort of calculus of similarity and a priori likelihood.

Word-level substitutions are familiar from the study of slips of the tongue. But a second observation about the errors of continuation in typing is that they seem rather different, in interesting ways, from word substitutions in speech. I'll stop for now, since my breakfast hour is over, with a quote from Thomas Berg, "Slips of the typewriter key", Applied Psycholinguistics 23(2) 185-207, 2002:

… slips of the key resemble slips of the pen, although not slips of the tongue. It is argued that speech errors are shaped by a fully deployed structural representation, whereas key slips arise under the influence of a weak structural representation. By implication, speaking is characterized by a hierarchical strategy of activation while typewriting is subject to the so-called staircase strategy of serialization in which activation is a function of linear distance. These disparate strategies may be understood as a response of the processing system to disparate requirements, such as varying speed of execution.

Berg is writing about a different type of slips — things like "rebember" for "remember" and "special droups" for "special groups". But I think that the quote is relevant anyhow — and I'll explain why some other morning.


  1. Arnold Zwicky said,

    May 22, 2008 @ 9:03 am

    There's a brief mention of "completion errors" in my posting on "broccoli rabbi" (and other things): http://itre.cis.upenn.edu/~myl/languagelog/archives/005392.html
    It could be a "completion error", a typo that results you start writing or typing a word and then drift part-way in to another word. I do this all too often with -ation and -ating words — starting the verb COOPERATING but ending up with COOPERATION, for instance. And several people have reported on the American Dialect Society mailing list that their intention to type LINGUISTS frequently leads them into LINGUISTICS, which then has to be truncated. (This discussion on ADS-L followed my typing "original Broadway case", with CASE instead of CAST, and commenting on it.)
    Since then:
    on ADS-L, 2/19/08: i did, and then i backtracked and worked out what much have been intended.
    2/23/08: in typing a Language Log posting, three times I aimed for “might be derided” but typed “might be derived” instead

  2. Rachel said,

    May 22, 2008 @ 9:20 am

    I do this a lot, especially when I'm typing a word that I type infrequently that starts the same way as a word that I type more frequently. I think it's similar to certain errors I make while playing the piano: my fingers are used to one sequence of movements rather than another, so that's what comes out when I'm not paying attention.

    Sometimes the words I type by accident aren't more frequent than the words I was trying to type, though. I used to not be able to type the word "hypothesis" because it would always come out "hypnosis."

  3. Chris Hunt said,

    May 22, 2008 @ 9:52 am

    I'm with Rachel – there's definitely an element of "muscle memory" in this. I'm a database developer, so there's words like "select" which I type *a lot*. Consequently there's a good chance of this kind of error when typing similar-starting words if fingers don't fully engage with brain.

  4. Marc A. Pelletier said,

    May 22, 2008 @ 9:57 am

    I would also tend to ascribe those errors to "mechanical memory", where the sequence of *movements* sidetracks into a more frequent pattern during production, rather than any sort of unconscious association.

    I know memory of movement sequences is very strong, and I remember reading a recent paper that showed with FMRI (saveur du jour) that once learned, a sequence of movement is produced from a qualitatively different part of the brain than when attention is being paid to a less familiar sequence — very possibly the source of the quip about never forgetting how to ride a bike.

    It certainly seems to me that my not infrequent substitutions of this sort are unfailingly in favor of a statistically much more frequent word, and always share the first few letters. This hints of slipping into a familiar mechanical pattern much more than selecting the "wrong" word. In other words, I'm quite confident that (at least for me) the substitution takes place at a very low level.

  5. rootlesscosmo said,

    May 22, 2008 @ 10:16 am

    Arnold Zwicky uses "drift" to characterize what happens part way through typing a word. Might we not hypothesize that when attention "drifts"–most likely to the next word, or the rest of the sentence–muscle memory takes over and produces this type of error?

  6. John Cowan said,

    May 22, 2008 @ 10:26 am

    On "normal" vs. "norman".

    Edgar Rice Burroughs, it seems, was reading a pulp magazine, and had one of those "I could do better than that!" insights, so he wrote down a story (the first of his Martian tales) and sent it to the editor under the transparent pseudonym of "Normal Bean" (i.e. someone of ordinary intelligence). The editor, however, read "Normal" as "Norman", and duly published the story under the byline "Norman Bean".

  7. Catanea said,

    May 22, 2008 @ 10:32 am

    Only lately, because it is raining, I seem ALWAYS to type rainging and have to go back and delete the mid g. Yes, it is the fingers, and not the thinking mechanism.

  8. Ryan Denzer-King said,

    May 22, 2008 @ 10:48 am

    I don't know a whole lot about speech perception or cognition, but could the error not be related to the overall phonological similarity in the words?

    On a related note, I've found that words I use more often tend to be represented more strictly phonologically in my mind (leading me to confuse their, there, they're in rapid typing such as IM), while less commonly used words are represented mostly orthographically, especially because in many of those cases I've never actually heard the word pronounced.

  9. Ellen said,

    May 22, 2008 @ 11:15 am

    For me, it's definitely a keystroke thing rather than a morphene thing, because it tends to happen to me with words that start with st-. I'll type the st, and then finish off with another work, usually one I frequently type.

  10. marie-lucie said,

    May 22, 2008 @ 11:17 am

    Practically all of my professional writing is in English, and when I write in French I can't seem to be able to type the word "langue" without first typing "language" and having to correct it. (I didn't this time, because I was paying close attention).

    I am also reminded of something which is perhaps similar: I was once in England with a group of French people, most of whom spoke little English. One woman would confidently start a sentence with one or two English words, then continue in French, apparently unaware of the switch. I remember the first time this happened, in a grocery shop where she wanted to buy some apples (you can tell how long ago this was, before self-service became general): "Give me … une livre de pommes", and the same type of thing happened a number of times in the following days. She did not seem to realize why people did not understand what she wanted.

    Yet another case I read about (sorry, I can't find a reference): a linguist was doing some dialect fieldwork in Greece, and his consultant was a very old lady who spoke quite fast. He kept urging her to speak slower, so she would say the first syllable of a word quite slowly, but then compensate by saying the rest of the word very fast. Another woman (who did not speak the dialect in question) would try to demonstrate to her what was meant, but to no avail.

    It seems that in all the cases reported (all involving sequential motions), the speaker's or writer's or piano player's conscious attention is concentrated on the beginning of the word, sentence or musical sequence and afterwards habit takes over. I wonder if this could be related to the fact that in languages with strong stress, unstressed components tend to lose their distinctiveness? or even more to the tendency of initial lexical components to attract stress?

  11. Ellen said,

    May 22, 2008 @ 11:22 am

    (work should be word. That one, however, was probably a typo — same finger, different hand.)

  12. Rosie Redfield said,

    May 22, 2008 @ 3:36 pm

    I'm making more and more of these errors (in typing), so I've been assuming it's an age-related thing. Has anyone looked for a correlation with age?

  13. Mark Eli Kalderon said,

    May 22, 2008 @ 3:44 pm

    "World's worst proofreader"…I doubt it. There's plausibly a cognitive asymmetry between proofreading your work and the work of others. My advisor, Paul Benacerraf, once explained it in terms of a difference between encoding and decoding. When proofreading your own work, what's to decode? After all, they are your thoughts. ;)

  14. felix culpa said,

    May 22, 2008 @ 4:09 pm

    Perhaps a related or analogous or utterly other matter; but still intriguing:

    In the mock-Cupertino address, the adjacency of the words “Log’ and ‘Plaza’ called up a vision of a shopping plaza built from Lincoln Logs (notched toy logs for building toy structures, and one of my favorite pastimes as a child).
    ‘Language Log’, if one were to pursue associations, suggests the classic
    Socratic non-ambulatory teaching medium, teacher at one end, student at the other of an archetypal log; but when ‘Plaza’ is added, the scale of the mind’s representation (or mine, anyway) leaps many orders of magnitude.

    Perhaps and inflection error, perhaps a weak structural representation, or some other error arising entirely elsewhere.
    Whatever the case, I find it striking.

  15. felix culpa said,

    May 22, 2008 @ 4:11 pm

    Ah, yes; ‘Perhaps an inflection…’

  16. Arthur Crown said,

    May 22, 2008 @ 4:44 pm

    Why, when I'm proofreading something, does stuff I've completely missed on the screen pop out on the printed page?

  17. marie-lucie said,

    May 22, 2008 @ 4:45 pm

    Proofreading: I find it much easier to proofread other people's work than mine: it is new to me, so I have to pay attention to every word. In rereading one's own work, especially work that has already been revised, there is nothing new and it is more difficult to pay close attention to every little detail. Sometimes it is only when the text is shown in another form – especially once it has been printed – that the occasional glaring error jumps at you.

  18. john riemann soong said,

    May 22, 2008 @ 4:51 pm

    "Has anyone looked for a correlation with age?"

    I'm only a high-schooler, so don't take my suggestion so seriously as I'm hardly as well-qualified as most of the commenters here, but I've been wondering if children view orthography (if not outright grammar) more "conceptually" than adults, who will tend to rely more on association. The reason I say this is based more on personal experience than any rigourous evidence (I'm just wondering about it). Some years back, I used to marvel at the disparity between the speed of constructing a sentence like, "Wouldn't it have been moving?", and the speed of consciously analysing the grammar of that sentence, in all its negative inversion conditional glory. The idea that I was following an association (as opposed to concept) that allowed me to spit relatively complex constructions out in fractions of a second occurred to me when I suddenly became tripped up when writing a nested hypothetical and no construction would magically appear in my head; evaluating which mood and tense to use suddenly became a real pain. When I was younger I used to have no such trouble, and automatically perceived a concept in every facet of language; I even perceived minimal pairs that didn't exist in English. For example I thought that the /n/ in "knight" was a slightly different consonant from the /n/ in "night" and I used to think the vowel was stronger in "night" than "nite". Currently, If I'm especially tired when burning the midnight oil, I might at times even absent-mindedly substitute "right" for "write," despite the fact that as a child I thought they were very different words, even in pronunciation, to the extent that when I was younger it would have been quite difficult for me to have made such a mistake. To me, association rules are rather like spell-checkers — they are not resource-intensive but they also bear a higher chance of error.

  19. Sili said,

    May 22, 2008 @ 5:04 pm

    I don't write professionally, but I have similiar experiences as marie-lucie. Most of my friendships and idle chat is conducted online in English, and that has been the case for a coupla years now. As a result my Danish typing is significantly slower and I'm very very likely to end a word in English or completely type out a word in English if the Danish and English words begin identically.

    The suggestion of this being a motor memory phenomenon makes me think that it would be interesting to see what sorta errors someone 'bilingual' in Dvorak and Qwerty makes. If such persons exist.

  20. Rick S said,

    May 22, 2008 @ 5:18 pm

    @Rosie Redfield: I learned to type at age 16, and have been a pretty fast typist ever since. But I used to hardly ever make this kind of mistake until about 10 or 15 years ago, when I was 40 to 45 years old, and at that time it became quite annoying. So I can see that there might be an age-related effect. On the other hand, backspacing is far easier and less time-wasting than using Wite-Out, so maybe it's just that it's way cheaper to type recklessly than it used to be in the days of typewriters.

    I've also noticed that my typos have changed over time. I used to type hte for the a lot (apparently, this is fairly common, though a Google search implies otherwise), and -tino for -tion, but I hardly ever do either of those any more. I suppose I've learned to be cautious when typing such words.

  21. Garrett Wollman said,

    May 22, 2008 @ 10:45 pm

    For quite a few years, I found myself unable to type "Cape Cod": it would always come out as "Cape Code". Now I seem to have moved on to another error (although I can't for the life of me remember what it is at the moment).

  22. dr pepper said,

    May 22, 2008 @ 10:49 pm

    I always had trouble with writing. My mother said i had a touch of dislexia, which she thought came from my father, who always had trouble with diagram. Personally i blame her for trying to make me switch hands just as i was learning cursive. Anyway, what used to happen a lot is i would start to write a lower case g or d and i'd end up writing the wrong letter, not because i was thinking of the wrong letter but because i'd make the tail loop in the wrong direction. I don't have that problem anymore since i switched to typeing.

    Also, when i'm driving a familiar route that can have 2 possible destinations, sometimes i'll be thinking of something else then suddenly realize i'm headed for the destination other than the one i was going to. I'm pretty sure that's a related error.

  23. Greg S said,

    May 23, 2008 @ 1:05 am

    In my cursive handwriting I sometimes write S instead of L or vice-versa, and I guess that this is because of their symmetry on the typewriter keyboard. That is, I'm guessing that my brain absorbed some S-L equivalence from typing, and then sometimes mixes them up when writing. I learned touch-typing in elementary school, and the mix-up happened more frequently when I was in high school (a couple of decades ago) than now; indeed I'm not sure it happens at all any more.

  24. Richard Sabey said,

    May 23, 2008 @ 2:33 am

    Rachel said "I do this a lot, especially when I'm typing a word that I type infrequently that starts the same way as a word that I type more frequently."

    That is true for me for that capture error which I have spotted most times in my own typing, namely "that" for "than". "at" for "an" is not especially frequent an error in other words, even though "an" is a frequent letter-sequence, so this is unlikely to be a substitution error.

    John Riemann Soong wrote 'If I'm especially tired when burning the midnight oil, I might at times even absent-mindedly substitute "right" for "write," '.

    I occasionally make such slips, e.g. "fare game" for "fair game" or "all though" for "all those". It seems that, sometimes, when I'm verbalising my thoughts, the words spring to mind in the form of their pronunciations, rather than their spellings.

  25. Barbara Partee said,

    May 23, 2008 @ 11:10 am

    One that happens a lot to me that I HOPE I've always caught, and I don't understand it because I don't think the substitute is one I use frequently: It's hard for me to start an e-mail with "Dear Chris" or "Chris!", because (only in the header!) about 10% of the time "Chris" comes out as "Christ"!
    And in place of straight -ing vs -ion confusions, more often my attempted -ion words come out as -iong. At least that's very recognizable as a mistake if I happen not to catch it.
    I also think there may be different errors when I'm freely composing vs. when I'm copying something. I remember lots of fast copying from the many years when I composed in longhand (all through the typewriter era and for a while into the computer age), and I got into a different mode when copying, practically fast-spelling to myself rather than thinking the words. Something more mechanical and very efficient. I could go really fast as long as I didn't stop and think. I have no idea what error-types were most frequent in that mode.

  26. TootsNYC said,

    May 23, 2008 @ 12:34 pm

    I vote "muscle memory" as a strong component.

    I have typed "there" when I *wanted* to type "their," and when I have consciously spelled out each individual letter: t h e i r — only to look at the computer screen and see "there."

    But I also think there's a "muscle memory" of some sort in the brain,

  27. john riemann soong said,

    May 23, 2008 @ 2:11 pm

    TootsNYC: I think I know what you mean! Eye movement patterns, perhaps, that begin to have a sort of conditioned/associated response to words, even with your eyes closed.

  28. John Cowan said,

    May 23, 2008 @ 3:48 pm

    I also have muscle memory for certain errors: for decades I have been typing "return" (a word much used by computer programmers) as "reutnr". Global search and replace is my friend.

  29. David Kutzler said,

    May 24, 2008 @ 4:51 pm

    I am fascinated by capture errors, simply because I am so vulnerable to them. The classic example of the capture error is when you are driving to the store on a weekend and suddenly find yourself pulling into the parking lot of your workplace. Because the route to the store coincided with part of the route to work, your brain was "captured" by the familiar driving-to-work behavior. I recall when my mother was refereeing an argument between siblings; my sister's arguments were being drowned out by my brother. At one point my frustrated sister raised her hand like she was in school to get Mother’s attention. She was quite embarrassed when she realized how silly she looked.

    Law enforcement agencies sometimes take advantage of capture errors to foil individuals whom they suspect of using a false identity. They have them repeatedly say aloud their name and social security number, and then subtly distract them. A surprising number of suspects will stumble and say their real name or social security number.

  30. Jon G said,

    May 26, 2008 @ 12:43 am

    @ Ryan Denzer-King
    My sense of this type o' typo (sorry, couldn't resist) is much more in line with yours; that of being similar in sound or somehow phonetically linked.

    I make as many slips, miscorrections and substitutions as the next person, but I have noticed (over the past two years especially) that when I type I will every once in a while, when slightly distracted, correctly spell a word that sounds similar (but never identical) to the word I intended.

    It doesn't seem to occur with any specific word more then any other, although usually both the wrong word and the word I was thinking of will be 'simpler' words.

    I know that I, when typing (which I do a lot of as my job involves lots of email), have a strong tendency to 'hear' the words I am typing in my head as I am typing them with a slight tendency to hear them just a step ahead of where my fingers are.

    I would also have to be characterized as a fairly verbal person, and I would put forward that the category of mistakes one makes more frequently could be in some ways linked to personality types. I could also be completely wrong on that, though.

    I can't, of course, think of an example off the top but I'm certain that I'll notice one or two shortly, now that I'm thinking about it.

  31. Bill Tozier said,

    May 26, 2008 @ 6:37 am

    I'm reminded strongly of the transcription errors that crop up frequently in closed captions for the hearing impaired, especially those that are typed "live". These substitutions can be egregious or subtle, but they're nearly all word substitutions of just the sort you describe.

    I note that I'm "hearing" these words I'm typing now, even as I frame and edit the sentences, and I know that sometimes my mind wanders as I remodel what I'm going to say on the fly.

    Might there be an aspect of transcription error, one that crops up when we're managing our own internal dictation?

  32. Rachael said,

    May 28, 2008 @ 12:40 pm

    I agree about the "muscle memory" errors. If I have to type "Katherine" it often comes out at "Kathering", because "ing" is such a common ending.

    Another category of typo I make is pre-empting a vowel or consonant cluster from the next word I'm planning to type, such as "goff off" where I mean "got off".

    "it would be interesting to see what sorta errors someone 'bilingual' in Dvorak and Qwerty makes. If such persons exist."
    I don't know anyone who uses Qwerty and Dvorak, but my husband is fluent in Qwerty and Fitaly, so I'll ask him.

  33. Aaron Davies said,

    May 31, 2008 @ 1:51 am

    Muscle memory is certainly part of it (I have a very hard time typing the name of my alma mater, Columbia, without a trailing ".edu") but I've noticed other things as well. One in particular seems to be a sort of typist's haplology–my most common typo for "remember" is "rember"; sometimes it seems as if the letters I know need to get typed simply get dropped somewhere between my brain and my fingers. (This happens occasionally with words too, as in the incredibly annoying phenomenon of people who unaccountably forget all the "not"s and "n't"s in their negative sentences. In my case, I have a very clear memory of the sentence I cast in my head ten seconds ago, but when I look on the screen, a word is simply missing.)

  34. Gregory said,

    December 18, 2008 @ 4:45 pm

    I have been a very fast (80 w.p.m. as verified by online tests) and accurate touch typist for twenty years. I think and speak extremely quickly. I'm turning 39 soon, and I've noticed in the last two years that my rate of four kinds of substitutions when typing has increased: (1) "sounds alike" substitutions (e.g. our for hour), (2) "different word with same initial characters substitutions (e.g. software for softer), (3) "think ahead" substitutions where a word later in the sequence seems to have interfered with characters in a preceding one, and (4) random substitutions with no discernible cause. Part of the problem seems to be that I think and type so quickly that the fingers are typing an earlier part of the sentence while the brain is already racing ahead to later words in the sentence. It's almost like the brain "fires off" a command to start typing "softer" and then stops paying attention and the rote practice of having typed "software" many times more frequently takes over and the fingers automatically complete the more frequent and familiar sequence. Hopefully this is just a normal aspect of aging. So now I try to give my emails a quick scan before sending, which is unfortunate since it slows me down slightly.

  35. Scott C said,

    March 25, 2010 @ 3:04 am

    I agree with everyone..since getting closer to forty I noticed I'm constantly substituting words for those that have similar sounds and starting characters. I also type fast and my brain is constantly thinking ahead to the next part of the sentence. I'm an 'adhd' type so I guess I get easily districted and the muscle memory takes over and takes the best guess for what it should type.

RSS feed for comments on this post