What is reading? One may engage in reading by way of seeing and comprehending a contentful inscription, of course, but that is simply a paradigm; it is not reading itself. Blind people, e.g., often read using Braille, and text comprehension does not seem necessary for reading. So, my project in this paper is to address this question: What is the proper analysis of person S reads text W? Surprisingly, no philosophical attempts to analyze reading exist; this question has (to my knowledge, anyway) yet to be tackled. Numerous psychologists, cognitive scientists, and linguists—Chomsky (1968) and Piaget (1952) are but two immensely influential examples one can point to here—have given a lot of thought to various crucial aspects of reading, but their specific interests always diverge from the narrow goal I have here. And of course, a great many philosophers have also set their sights on the activity of reading and attendant issues. Worth (2017), e.g., provides a thorough defense of the moral and social value produced by reading; Heidegger, Derrida, Strauss, Ricoeur, and countless others working in a primarily continental tradition have contributed an enormous amount concerning text exegesis and hermeneutics.

My aim here, however, is one of pure conceptual analysis. While progress on the theoretical matter could pave the way for progress elsewhere—e.g., informing discussions in epistemology regarding the nature of testimony and aiding debates in esthetics regarding emotional responses to fiction—I leave those applications to others.Footnote 1

“Reads” is, unsurprisingly, ambiguous. There are senses of “reads” on which text comprehension is not a requirement. Malcolm can truly say he’s read The Handmaid’s Tale even if he has not grasped all of its content; Angus can truly report that he read the road sign, even if he falsely believed that it said speed limit 100 km/h. Fred may even read a sentence of German without ultimately understanding a single word (more on this in the next paragraph). But there is also a sense of “reads” that requires that some cognitive success has been achieved. If Lana were to visually take in the sentence “All tigers are mammals,” and then she sincerely reported its content to be some numbers are prime, one could reasonably claim that Lana must have failed to read the sentence. (And metaphorical senses of the term exist also. One may say, “The midfielder read the defense superbly,” “The comedian quickly read his audience,” or “Janine failed to read the warning signs for skin cancer.” While these sentences are true only if some proper sort of understanding genuinely occurred in the mind of the midfielder or the comedian, or failed to occur in the mind of Janine, none of them deal with the factive sense of “reads” just outlined, nor any sense of the term that is the target of this paper.)

I am here interested in “reads” only in the senses exemplified by the first few examples (viz., the Malcolm, Angus, and Fred examples) discussed above. While focusing on the senses of “reads” that do not strictly imply any sort of text comprehension, it is helpful to make some finer distinctions. Cases of these kinds of reading may be exemplified in either of the following two ways: (i) person S engages with text W, but W lacks linguistic content altogether, or (ii) person S engages with text W, W is contentful, but nonetheless S fails, for whatever reason, to map W to any of its content. To illustrate (i): suppose Alex is at the optometrist’s office and is asked to read the eye chart in front of him. If he sincerely and accurately reports that he is seeing “A” followed by “E” followed by “O,” then he has read the chart, but this succession of letters has no linguistic content on its own. Or suppose Eddie sees “flamgorggshy” tokened on a page. The inscription has no content in English (or any other language), but Eddie, familiar with the English alphabet, is able to read it (to put it roughly, he is at least able to mentally “sound it out”).

To illustrate (ii): consider our mono-linguistic English reader, Fred, who visually encounters the contentful German inscription, “Schnee ist weiss.” Fred’s ability to recognize enough of the letters of the German alphabet will allow him to at least mentally “sound out” the inscription. This is enough for reading to have occurred on this occasion. Alternately, the above examples concerning Malcolm’s inability to map some of the inscriptions he encounters in The Handmaid’s Tale to any content whatsoever, or Angus’ believing the road sign to say speed limit 100 km/h (when the inscription on the sign is in fact “Next Rest Area: 100 km,” say) serve to illustrate (ii). The sense of reading captured by (ii) is what we may call reading-in-language-x (in short, readingL). Fred may be unable to readL any sentence or even any individual word of German, and thus be reading-illiterateL with respect to German, while not being reading-illiterateL simpliciter. One who is reading-illiterateL with respect to every language is reading-illiterateL simpliciter. To simplify, I will for the rest of this paper limit my discussion to the proper analysis of readingL (but I will henceforth just use “reads” and “reading” to express this concept).

Again, a paradigm case of reading—visually attending to a contentful inscription—is not reading itself. Many blind people are capable of reading by way of feeling patterns of raised dots. But if vision and touch may facilitate reading, it seems chauvinistic to deny that reading using other sensory modalities may occur. Are there principled reasons for thinking that someone who has heard the audiobook version of The Handmaid’s Tale, e.g., has failed to read that story? And if texts may be read visually or tactilely or audibly, what reason could there be for thinking that reading couldn’t occur using taste buds or olfactory receptors? These would be exotic scenarios, but provided information was properly encoded, i.e., provided the relevant (sequences of) flavor tokens or odor tokens expressed linguistic content, it seems as though we cannot rule out as metaphysically impossible the tasting or smelling of a text.Footnote 2

So, here’s an initial attempt at an analysis of reading:

FormalPara Proposal 1

Person S reads text W if and only if S uses at least one sensory modality to cognitively engage with W’s content (that is, the word meanings, meanings of phrases, or the propositions semantically expressed or pragmatically conveyed by W).

“Cognitive engagement” does not imply accurate, complete, or even partial comprehension on any one occasion, so misreading a text still counts as reading on this analysis. But the activity of reading essentially involves cognitive engagement; a mindless computer or a thermometer may be said to engage in “reading,” but that should not count as genuine reading. In order to read, a mind must attend to a text, although one’s attention may be fleeting or cursory. But to capture the sense of reading-illiteracy Fred (from above) may have with respect to German, we should recognize that the sort of cognitive engagement required to for reading literacy involves more than the sort of mental processing needed to at least mentally “sound words out.” The sort of cognitive engagement required must involve attempts to comprehend the content of texts that are in fact (non-accidentally) successful on at least some occasions. Generally speaking, someone who relentlessly fails for whatever reason to comprehend the content of any of the texts she encounters counts as reading-illiterate. More specifically, Fred, while being able to at least mentally “sound out” “Schnee ist Weiss,” should not count as being able to read the sentence if he relentlessly fails on every occasion of his interacting with tokens of the sentence, his best efforts notwithstanding, to comprehend those tokens.Footnote 3

(Moreover, given what was noted above regarding reading that does not occur using vision, “text” cannot mean inscription. Soundwaves, sequences of tastes, and odors all fail to count as inscriptions, yet since each of these things seem at least metaphysically possibly read, we need a broader understanding of “text.” So, let us understand “text” as any meaningful word token, phrase token, sentence token, or collection thereof.Footnote 4 And while the ordinary sensory modalities of Homo sapiens can facilitate the reading of a text, we should further admit that there may be alien people having alien sensory modalitiesFootnote 5 capable of facilitating reading.)

So, the following modification to our current proposal is in order:

FormalPara Proposal 2

Person S reads text W if and only if S uses at least one sensory modality to cognitively engage with W’s content, and S does not relentlessly fail to grasp the entirety of W’s content.

Proposal 2, however, is faulty, and the following, fairly obvious objection seems decisive. Suppose I were to have a daydream whose content turned out to be the content of some text W, or perhaps the entirety of some text W’s content gets implanted in my mind thanks to hypnosis or computer download. Or suppose I simply choose to transcribe the contents of a story that I recently thought of. In such cases, I may be cognitively engaging with some content, and that content happens to be the content of some text, but even if I encounter that text visually, say, I haven’t thereby read that text. This tells us that the accidentality or infelicitousness of a reader-text connection must be eliminated in a proper analysis of reading, and some appropriate priority of text to reader must be accounted for.

The most natural way to meet these conditions is to require an appropriate sort of causal connection between text and reader, such that S’s engagement with the content of W is ultimately brought about by W itself. Of course, other factors besides the mere presence of W must be in place for S to cognitively engage with W (e.g., S must possess some appropriate sensory modality for the sort of text she is in the presence of, S must be in an environment that lends itself to the use of that sensory modality, etc.), but this sort of requirement will rule out cases in which the cognitive engagement is inappropriately related to the text in question or is itself (partly) responsible for producing the text.

So, here is a revised proposal:

FormalPara Proposal 3

Person S reads text W if and only if S cognitively engages with W’s content using at least one of S’s sensory modalities, S does not relentlessly fail to grasp the entirety of W’s content, and S’s cognitive engagement with W is (at least partly) caused by W (and not vice-versa).

This analysis meets all of the above desiderata. Misreading on an occasion may count as reading; reading can be facilitated by multiple sensory modalities; mindless objects cannot read; texts are responsible for producing episodes of cognitive engagement, and not the other way around.

A fairly compelling duo of objections to Proposal 3, however, focuses on the alleged liberality of the analysis.Footnote 6

Objection from Hearing, Take One: Hearing is not a sensory modality that can play any part in facilitating reading; a text, in whatever form it takes, is a thing that is complete and fixed at the time of engagement, but when someone is listening to words being spoken, this medium is neither complete nor fixed at the time of engagement. Audio, especially in the form of verbal output, may even be wildly extemporaneous. Such is not the case for the paradigm cases of readable texts—inscriptions and the raised dots used in Braille.

Reply: Some sequences of soundwaves—recorded audio—are just as complete and fixed as anything that has been inscribed or displayed as a pattern of raised dots. Moreover, inscriptions or patterns of raised dots, and texts generally speaking, need not be fixed nor complete. We could easily imagine a written text, e.g., that unfolds in extemporaneous or unpredictable ways. Suppose a computer makes random word tokens sequentially appear on a monitor. One may not have any idea what’s coming next, but reading could still be occurring.

More positively, one may reason as follows. Suppose one has (only) heard The Handmaid’s Tale (perhaps in audiobook form). Such a person may understand exactly the same content as one who has seen an inscription of the sentences that comprise the story, her level of proficiency with the content of the story may be exactly the same, the content may affect her emotions in the same ways, and she may be disposed to offer similar literary, moral, or social critiques of the story. There is thus nothing relevant to the activity of reading that distinguishes the cases. Hence, one who has (only) heard the story has read the story.Footnote 7

Objection from Hearing, Take Two: Hearing is not a sensory modality that can play any part in facilitating reading; if it were, then not only could listening to a recorded audiobook count as reading, but so could listening to a song, hearing a lecture, watching a non-silent movie, watching a non-silent play, and listening to a live, word-for-word re-telling of The Handmaid’s Tale. These results are intolerable.

Reply: These results do indeed seem intolerable. This suggests the following:

FormalPara Proposal 4

Person S reads text W if and only if S cognitively engages with W’s content, S does not relentlessly fail to grasp the entirety of W’s content, S’s cognitive engagement with W is (at least partly) caused by W (and not vice-versa), and listening is not the sole sensory modality S uses to engage with W’s content.

Patrick Fleming has defended Proposal 4 and has argued that the reason why these cases are in fact intolerable is because each involves some agent other than S actively mediating S’s interpretation of W. Reading, he insists, is an activity that requires a direct, unmediated relation between reader and text, one where S is autonomously making judgments about W’s content, such that “all the interpretative questions in the text are being decided by the reader alone.” Fleming goes on:

The difference between reading Hamlet and seeing the play is that the actors make choices about the text that is reflected in their performance. Reading is a direct engagement with a text. Even [listening to] an audiobook is not reading because the person on the tape is making choices about how to interpret the text. They may change their tone of voice or pause to linger on a word. In that case, one is listening to a reading, not actually reading. A reader needs to directly wrestle with the text, and this cannot be assisted by cues from others. Reading cannot be mediated by someone else’s performance or interpretation. (personal correspondence)

The criticism here strikes me as off the mark. There may be cases of reading where S is simply unable to answer all interpretative questions concerning some text W. S herself may even realize that she lacks this ability. S may find herself at sea about how to interpret some parts of a text she has in fact read—think about difficult pieces of poetry or cryptic song lyrics. Moreover, one’s decisions concerning which contents are expressed by W may indeed be mediated in important ways by the actions and attitudes of other agents; perhaps S needs to seek out some assistance in interpreting W, knowing she lacks the cognitive tools to arrive at some legitimate interpretation of W completely on her own. And a purely autonomous development of some legitimate interpretation of W might simply be beyond S’s abilities—think about the assistance parents or teachers must give novice readers. Relentless failure to arrive at some legitimate interpretation is what must be ruled out in an analysis of reading, but as long as some legitimate interpretation of W is ultimately grasped, it simply doesn’t matter whether or not S’s grasping has been mediated by the actions or attitudes of others.

If Fleming’s suggestion is furthermore intended to suggest a sufficient condition for reading, other unfortunate results may be in the offing. Suppose an unintelligent computer were to produce a sequence of sounds, one indistinguishable from an ordinary, human-produced song or audiobook, say. A person S may engage with these sounds directly, and then autonomously arrive at an interpretation of those sounds. Such an interpretation, we may further suppose, could be one that is unmediated by any other agent in any salient way. Heard songs, heard audiobooks, etc. are precisely the sorts of cases Fleming means to take care of, but his suggestion may imply, contrary to his intuitions and my own, that reading occurs in the sorts of cases just described.Footnote 8

At this point in the discussion, one may put forth this complaint: if Proposal 4 is correct, what then is the principled difference between the uses of other sensory modalities and this one particular sensory modality of hearing that cannot alone facilitate reading? Why does it seem chauvinistic to exclude other ways of interacting with a text yet not this one? (And what about alien physiologies with as yet unknown to us sensory modalities?) What justifies acceptance of this sort of liberalism, a sort that condemns only hearing?

In the end, I think that Proposal 4 cannot be defended in a non-ad hoc way. Ultimately, something very close to Proposal 3 seems correct; any sensory modality may in principle facilitate reading, even hearing. What’s needed is an analysis that relies on a more detailed account of the nature of contentful texts and what exactly it is that we cognitively attend to when we read them.

Merely seeing tokens of contentful words or phrases (even given an ability to at least mentally “sound them out”) isn’t sufficient for reading. There must be an attempt to grasp the content of those words and phrases using one’s vision, and repeated attempts cannot relentlessly result in failure. But when one makes such an attempt, what’s absolutely crucial is that in so-doing one pays attention to word structure. One who can read inscribed texts using vision is able to do so precisely because she is able to attend to word structures, e.g., spellings of English words or stroke patterns of Chinese words, and this is what’s facilitating her grasping (when it occurs) of the contents of those words (and ultimately for the grasping of the content of whole phrases formed from those words). This is a paradigm case of reading, but what’s essential to reading itself is this attendance to the structural building blocks of words so as to grasp content. Now, readable English or Chinese words are usually represented by collections of inscribed letters or stroke patterns rather than collections of phonemes (or odors or tastes), but this is not necessarily the case.

Spoken words and inscribed words are fundamentally different sorts of linguistic items. Spoken words (likely) came first in human development, and competency with spoken words (normally) comes before competency with inscribed words in individuals of our species. A spoken language may exist without any of its words ever being inscribed, and illiterate people still may be competent (spoken) language users. When word inscriptions eventually arrived for our species, and when they eventually arrive for individual humans, such inscriptions serve as surrogates for spoken words (even though some literate language users may be utterly mute). The discrete parts of word inscriptions, i.e., letter inscriptions or stroke inscriptions (or any squiggle-squoggle, hieroglyph, or wisp-of-smoke inscription, for that matter), are ultimately founded on phonemes. Paradigm examples of reading involve an exhibition of a developed cognitive skill, an ability to attend to inscribed words and their inscribed parts, so as to grasp content. But were there collections of phonemes that were instead founded on word inscriptions, one would then have an audible text that could be read by people possessing the appropriate cognitive skill.

Suppose, for example, that someone became very adept at listening to someone spell out stories in English (“‘o’ ‘n’ ‘c’ ‘e’ ‘u’ ‘p’ ‘o’ ‘n’ ‘a’ ‘t’ ‘i’ ‘m’ ‘e’ ‘a’ ‘u’ ‘n’ ‘i’ ‘c’ ‘o’ ‘r’ n’…”). The spellings, the word structures, would be her primary cognitive focus, even though the medium of delivery would be soundwaves. It seems to me that would be a case of reading, and it would count precisely because she is attending to word structure so as to grasp content. In such a case, the words are not audibly washing over the person wholesale as spoken words usually do; it’s the attendance to their very structure that is playing the key role in their comprehension.Footnote 9 In this case, the phonemes are serving as surrogates for inscriptions.

Imagine that there were an alien species that developed differently from us Earthlings. Suppose in the development of an ancient Venutian species, and in the development of individual ancient Venutians, written words came first, spoken words later. Venutians (both as a species and in individual cases) eventually took their words-as-sequences-of-inscribed-squiggles and encoded them as phonemes. Encoding inscriptions as phonemes was an acquired skill for them, part of what it was for a Venutian to be considered literate. Moreover, Venutians who could not properly speak or hear such contentful phonemes were considered illiterate, even though many of them were perfectly competent users of the Venutian (written) language. Paradigm cases of reading for Venutians were instances of audition; how to attend to phonemic word structure to grasp linguistic content was a skill taught to young Venutians who had much earlier in their lives learned to write the Venutian language. (We could imagine a similar story being told for a developmental move from inscriptions to collections of smells or tastes.)

The upshot of this gedankenexperiment is that words with structures capable of facilitating comprehension can in principle be represented as soundwaves (or odors or tastes or wisps of smoke or pictures, for that matter). Reading primarily concerns the ability to attend to linguistic structure at the level of words, no matter how such structure is represented. Such structures are the basic units of reading comprehension. One can see and touch a text without reading it provided one is not attending in a primary sort of way to word structure. In listening to an audiobook, hearing a lecture, listening to a movie, watching a play, or listening to a song, one is usually letting the words wash over her wholesale. This is something that is difficult (perhaps humanly impossible) to do when reading a text visually or tactilely. And that’s why visual reading and Braille reading are paradigm cases of reading! One is forced, in a sense, to attend to word structure in such cases. But someone with proper training (of our species or another) in the presence of properly encoded texts, i.e., contentful texts where the word structures are cognitively accessible, could very well read using audition (or even smell or taste). She would just have to utilize the particular auditory skill she developed over time to do so (although I suppose it’s not impossible that such skills be innate), that is, deploy her facility with attending to embedded phonemic word structures for the purposes of grasping content.

Objection from Hearing, Take Two is thus not compelling, and Proposal 4 is left unmotivated. A refined version of Proposal 3, the proposal I ultimately think is correct, runs as follows:

FormalPara Final Analysis

Person S reads text W if and only if S uses some sensory modality for the primary purpose of cognitively attending to the word structures embedded in W for the purpose of grasping the content of W, S does not relentlessly fail to grasp the entirety of W’s content, and W is (partly) causally responsible for S’s use of her sensory modality for that purpose (and not vice-versa).Footnote 10