I recently installed a speed reader on my tablet. Now I wonder how the minimal time needed to read an $N$-letter word grows.

There was this previous question on the relationship between sentence length and reading time. However, I'm particularly interested in the relationship between word length and reading specific words.

Does the time required to read an $N$-letter word grow linearly?

  • 1
    $\begingroup$ You don't actually read letter by letter. You recognise familiar words by some key letters (usually the first) and their overall length and shape (ascenders and descenders), and you infer words from the context (you expect certain words in certain places). The length of the word has no bearing on reading speed. $\endgroup$
    – user3116
    Feb 5, 2014 at 22:23
  • 1
    $\begingroup$ I think this is a duplicate of How long does it take to read X number of characters?. That question specifically discusses linear, versus sub-linear, versus super-linear growth. $\endgroup$ Feb 6, 2014 at 6:07
  • 2
    $\begingroup$ I think the previous question was focussed on reading sentences and this one tended to have an emphasis on words, which I think is a small but relevant difference. I've tweaked both to hopefully prevent this one being closed as a duplicate. $\endgroup$ Feb 6, 2014 at 6:36

2 Answers 2


The average response times for single words in a lexical decision task (depart real words from nonsense words) is plotted below.

The curve follows a U-shape and span over several dozens of milliseconds.

enter image description here

Figure 2. Average reaction time and 95% confidence interval for words with lengths from 3 to 13 letters if length was the only factor having an influence (all other factors have been partialled out).

Ref: New et al., PBR (2006). Reexamining the word length effect in visual word recognition: New evidence from the English Lexicon Project

  • 1
    $\begingroup$ That is a consistent finding in many studie, but that figure shows the results from a lexical decision task. I explained in my answer why that is not relevant to the question: discerning words from non-words is not something you do while you read a text that only contains words. $\endgroup$
    – user3116
    Feb 6, 2014 at 15:59
  • $\begingroup$ @what I understand your point. We could also argue that the naming task reflects more the naming part than the reading one. $\endgroup$
    – marsei
    Feb 6, 2014 at 21:17
  • $\begingroup$ Since there is usual subvocalization in reading, if you don't suppress it, naming is much closer to reading than lexical decision. $\endgroup$
    – user3116
    Feb 6, 2014 at 22:01
  • $\begingroup$ @what in my view, a word can be seen either as a visual object with linguistic properties (studied in lexical decision) or as a linguistic object with visual properties (studied in naming). Both approaches refer to words and reading. In the case of a lexical decision, the reading ability reflects more the visual recognition part of reading, while naming (i.e. reading aloud) emphases the phonological/articulatory part of reading. In this light, there is a word length effect in (the visual part of) reading but this effect don't show up a naming task. My guess is that we could agree on this, no? $\endgroup$
    – marsei
    Feb 7, 2014 at 9:16
  • $\begingroup$ Lexical decision only asks for recogntion, naming asks for recall. When you read, you need to recall the meaning of the words your read, that is why in my opinion LDTs reflect what goes on in reading significantly less than naming tasks, where only the additional speaking differ from silent reading (where most of us usally subvocalize: en.wikipedia.org/wiki/Subvocalization). $\endgroup$
    – user3116
    Feb 7, 2014 at 11:45

To answer your question, we must first understand what you mean with "read". Since you ask in the context of speed reading, you appear to be interested in extracting meaning from written text. So we can rephrase your question as:

Does the time to recognize and understand a word increase with word length?

The effect of word length has been studied for two tasks, lexical decision and naming. In a lexical decision task participants are presented with words or nonwords (e.g. "dftrpg") and have to quickly press one specific key for words, and another for nonwords, on the keyboard. The time between the presentation of the stimulus and the key press (the reaction time) is registered. Since you are reading a text in which there are no nonwords, lexical decision is irrelevant in the context of your question.

A naming task is similar to a lexical decision task, with the exception that participants read the words aloud, and their responses trigger the computer via a microphone ("voice key"). Again, the reaction times between the presentation of the stimulus and the spoken word are registered. Since we cannot directly measure understanding, we assume that naming reflects understanding. (Of course you can speak words that you do not know, and even nonwords, but the reactions times would be significantly longer, so we would notice the difference. As long as RT is reasonably short, we may probably assume that the participant knows that word and has understood it.)

Not every time a participant names a word will he or she do so in the exact same time. Sometimes someone will be a bit slower to react, sometimes a bit faster. So we calculate the average reaction time for each word. What we note is that these average reactions times are not the same for different words, that is: the reaction times vary. Now, what we want to do, is explain this variance. Is it in some way related to word length? Or does it correlate to other characteristics of the words, such as number of syllables, word type (noun, verb, adjective), word frequency (how often a word is actually used in everyday spoken language), and so on.

After a number of calculations on the data, we find that all these aspects explain a part of the variance in reaction times (Baayen, Feldman & Schreuder, 2006). Here is a table of the different word characteristics (the predictors), and how much of the variance in naming speed they predict. Only those predictors are listed that had a significant effect. The numbers are adjusted R2 × 100, i.e. the percentage of variance uniquely accounted for by each predictor over and above all other predictors. You don't have to understand the meaning of these numbers, but they show the relative strength of the effect each predictor has on naming speed.

Lower bounds for the percentage of variance (adjusted R2 × 100)
explained by the significant predictors in word naming

Frequency                           6.08
Initial diphone 2                   4.33
Initial diphone (syllable-based)    3.88
Frication/Length first phoneme      2.24
Voicedness first phoneme            1.73
PC2 Consistency                     1.28
Length                              1.14
PC1 Consistency                     0.91
Written-to-spoken ratio             0.59
Neighborhood density                0.36
PC3 Consistency                     0.36
Inflectional entropy                0.34
Residual familiarity                0.31

In all, the results are:

  • The overall variance in reaction times is small. In young adults they range from around 300 to around 700 ms, with about two thirds of the RTs falling into a narrow slot of only 100 ms.
  • All the predictors together explain only half of the variance in reaction time (R2 = 0.4999), the rest of the variance must be attributed to unknown predictors or, more likely, are artifacts of the testing situation (distraction, boredom, etc.).
  • The single strongest predictor of naming speed is word frequency.
  • The contribution of each of the other single predictors is tiny (a few microseconds each).
  • Word length predicts only a few microseconds in variance.
  • There is a word length effect only if words are presented individually, not in a text context (Meyer, Roelofs & Levelt, 2003).

To answer your question, we can say that the length of the words has probably no noticeable effect on your text reading speed.


  • Baayen, R. H., Feldman, L. B., & Schreuder, R. (2006). Morphological influences on the recognition of monosyllabic monomorphemic words. Journal of Memory and Language, 55(2), 290-313.
  • Balota, D. A., Cortese, M. J., Sergent-Marshall, S. D., Spieler, D. H., & Yap, M. (2004). Visual word recognition of single-syllable words. Journal of Experimental Psychology: General, 133(2), 283.
  • Meyer, A. S., Roelofs, A., & Levelt, W. J. (2003). Word length effects in object naming: The role of a response criterion. Journal of Memory and Language, 48(1), 131-147.
  • $\begingroup$ What? (pun ;-) How does "the length of the words has probably no noticeable effect" relate to MacDuff's answer? $\endgroup$
    – draks ...
    Feb 6, 2014 at 15:44
  • 1
    $\begingroup$ The plot in MacDuff's answer shows the results of a lexical decision task. I explained why that is not relevant to the question: discerning words from non-words is not something you do while you read a text that only contains words. $\endgroup$
    – user3116
    Feb 6, 2014 at 15:58
  • $\begingroup$ To add to the debate, we can argue that naming (i.e. reading aloud) is not something you do while you read silently. The reading processes measured in the naming response times reflect too much the articulatory part of reading aloud. But there is no task that reflect pure reading - I replied to @what in a comment of my answer and tried to be more consensual though. $\endgroup$
    – marsei
    Feb 7, 2014 at 9:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.