As you are reading this, even though you are reading silently, you may experience the sensation of hearing your voice speaking the words in your head – you may even be moving your lips. It was originally believed that because we learn to read orally before learning to read silently, this inner speech is merely an epiphenomenon that has no real functional significance and is a carry-over from how we initially learnt to read. It, however, seems unlikely, given its pervasive nature, that inner speech has nothing to do with understanding written text and this would suggest it serves some purpose for reading.
It has been suggested (Gibson, 1965) that reading is little more than speech made visible and as we hear children sound out words, and we know that children tend to read only after speech has developed well, this may seem plausible. However, something more profound would appear to be going on as we read silently (300 wpm) far quicker than we read aloud (150 wpm) (Rayner and Clifton, 2009). So, the inner speech of skilled readers appears to be somehow compressed, but this compression is not simply reserved for difficult words, as we appear to be aware of it throughout the entirety of our silent reading experience. Indeed when we read a letter from someone we know, we appear to ‘hear’ their voice, including their accent, in our heads (Brown, 1970). Furthermore, if we read something written by someone who speaks quickly, we read more quickly (Kurby et al., 2009).
Inner voice appears to have two elements. First, there is the phenomenon of subvocalising, whereby the muscles in the speech tract are activated during silent reading. Second is phonological coding which refers to the mental representations of speech that gives rise to the experience of hearing sounds in our heads but will not always lead to conscious experience – we don’t always notice it; indeed when we do, it can disrupt our reading.
Let’s try a little experiment to see if we can separate the two phenomenon. Try reading the next few lines whilst say (out loud) ‘blah, blah, blah’ continuously as quickly as you can. The result may have surprised you in that you can easily hear the voice in you head whilst your mouth was fully engaged with something else. So, the voice in your head (phonological coding) is not dependent on subvocalisation (movements in your speech tract). More skilled readers appear to utilise subvocalisation less than poor readers (Edfeldt, 1960) but there does appear to be significant benefit to it, as it is utilised by readers during more complex passages of text (Hardyck et al., 1970). When subvocalisation is eliminated entirely, comprehension of complex text appears to suffer.
Try reading this silently:
The buoy and the none tolled hymn they had scene and herd a pear of bear feat in the haul.
And this…
If yew kann sowned owt thiss sentunns, ewe wil komprihenned itt.
You will almost certainly have found that you slowed to ensure meaning.
If inner speech were not a phenomenon, text that has been altered but is still pronounced correctly should be no harder to read than unaltered text when reading it silently…but it is. And tongue twisters – Barbara burned the brown bread badly - are also read more slowly when read silently (Haber and Haber, 1982). Furthermore, the evidence suggests that profoundly deaf children struggle to learn to read (Treiman and Hirsch-Pasek, 1983) – it should be noted that some profoundly deaf children (about 4.5%) do manage to learn to read well (Conrad, 1977) and those with profoundly deaf parents learnt to read better than those with hearing parents (Hoffmeister, 2000). Additionally, words that take longer to pronounce, take longer to read silently (Rayner et al., 2011).
So, if as the research suggests, the sounds of words influences the speed and accuracy of silent readingwe must conclude that inner speech is important to reading (Rayner et al., 2011). But is we are hearing the sounds in our heads, why can we read silently faster than orally. One reason appears to be that when we read orally our voice lags behind our eye movements by about half a second. This time lag is much shorter when we read silently (Rayner et al., 2011) - (more on eye movements here). A further reason may be that the phonological code is abbreviated when reading silently – this is identical to when you ‘hear’ yourself think suggesting that silent reading is externally guided thinking (Neisser, 1967; Rayner et al., 2011).
The importance of inner speech to reading appears to have two interlinked elements. First, phonological coding (the voice in your head) enables us to hold words and word order in our working memory – so by ‘hearing’ the words, they are easier to remember. This is held here until meaningful units can be passed to long-term memory. Because sentences are often long and the distance to the creation of meaning far, holding information as phonological codes supports working memory to cope with meaning extraction – this chimes with research that suggests the biggest cognitive load when reading occurs at the end of a sentence, when the meaning is fully extracted (Lobina et al., 2017).
This capacity to support working memory facilitates reading beyond mere memory of words but enables the creation of prosody during silent reading – vital for comprehension (Frazier et al., 2006; Carlton, 2009) – more here on prosody. Slowiaczek and Clifton (1980) maintained that prosody was a requirement for the comprehension of a connected discourse and Fodor (2002) hypothesised that in silent reading the prosodic contour projected onto the words directly affected the resolution of meaning – in other words, what we ‘hear’ in our heads, is the meaning. The implication is clear. The more prosodic the reading, the greater the likelihood that we understand what they have just read.
This implies that the development of prosodic reading should be a significant factor in fluency instruction – more here.
Does that suggest that oral reading may be better for comprehension? This is often suggested by reading experts but should come with a warning. Reading out loud is associated with improved memorisation and affords a production benefit, but not comprehension (Roberts et al., 2024) – where questions were a close match to the information read there was a benefit to reading the text out loud. This should not be a surprise, if as Kahmi (2007) suggests, reading comprehension is not a skill, but a complex mix of thinking, reasoning, imagining and interpreting and is a largely knowledge-based process dependent on content knowledge – more here. So it would be good advice to afford poorer readers the facility to read aloud in a comprehension test, but it will probably only help them with retrieval questions.
Follow The Reading Ape on Blue Sky @thereadingape.bsky.social
Comments