Words are also in the Western world recognized

as a whole - visually.

http://www.eurekalert.org/pub_releases/2009-04/gumc-bpw042409.php

http://www.eurekalert.org/pub_releases/2011-11/gumc-srr110311.php

There is also a rumor about an English study.

According to it isn't necessary how the letters

are permuted as long the first or last letter are

at their original position. No reference, note

it's a rumor. But the effect seems to exist.

Related:

http://www.eurekalert.org/pub_releases/2006-04/cp-ba041406.php

http://phys.org/news/2012-07-key-word-words-text-keywords.html

http://www.eurekalert.org/pub_releases/2008-11/afps-wci112508.php

http://www.eurekalert.org/pub_releases/2009-10/wt-cgh101309.php

http://www.eurekalert.org/pub_releases/2011-02/thuo-022211.php

Regards,

Joachim

Thanks for a quick response. Whole words are bad candidates to be units of reading. Of course, some short and frequent words we perceive as whole units. But foveal visual field is usually only 4-5 letters wide, many words are longer. New words, including names, do not exist in our mind, we have to perceive them by parts - letters, morphemes or syllables. And one more reason is the rumor you've mentioned. It is not a rumor, it is a fact. We can perceive words with letters mixed within the edge letters - it is called "edge effect" (Stevens, Grainger 2003; http://ureca.recherche.univ-lille3.fr/sparrow/ter/0607/stevens_grainger.pdf; Jordan et al. 2003; http://psycnet.apa.org/journals/xlm/29/5/883/) and "transposed letters effect" (Schoonbaert & Grainger 2004; http://sites.univ-provence.fr/wlpc/pagesperso/grainger/pubpdf/Schoonbaert&Grainger04.pdf). The ability to correctly perceive misspelled words means that not whole words are units of reading.

Maybe there is not such unit, but several of them that we process in parallel. As soon as one of them reaches certain level of activation enough to solve the reading problem, the network cools down to allow the next unit to enter. Something like the multiple read-out model (Grainger & Jacobs, 1996).

Vision is the quickest sense; it does recognize anything and deals with reading in a single and sudden way, just as a reflecting mirroring. That is why one can read several pages into very few seconds through dynamic reading.As I told you before the vision radar. There is a limit for edges into transposed letters.The read out model relies on the words already read in. If you read a book into another language you may feel that one or another word will be not understood, but due to the paragraph, you tend to choose one possible meaning just because you had not read that word before. So, as more you read more you can read into any language. I agree with Paloma, no units to understand what you read.Imagine writing and its relation with units. remember the way we start reading as children, understand that the units are real, but as soon as one starts reading well, no more units.

The critical point to read into Russian for us is to adopt another different and unrecognizable alphabet, but as soon we hear something can be understood.

if yuo can raed tihs, you hvae a sgtrane mnid, too.

Can you raed tihs? Olny 55 plepoe out of 100 can.

i cdnuolt blveiee taht I cluod aulaclty uesdnatnrd waht I was rdanieg. The phaonmneal pweor of the hmuan mnid, aoccdrnig to a rscheearch at Cmabrigde Uinervtisy, it dseno't mtaetr in waht oerdr the ltteres in a wrod are, the olny iproamtnt tihng is taht the frsit and lsat ltteer be in the rghit pclae. The rset can be a taotl mses and you can sitll raed it whotuit a pboerlm. Tihs is bcuseae the huamn mnid deos not raed ervey lteter by istlef, but the wrod as a wlohe. Azanmig huh? yaeh and I awlyas tghuhot slpeling was ipmorantt! if you can raed tihs forwrad it.

The unit of reading is different for children and adult readers. Again here, there are lots of differences according to orthographic transparency. I mean If the writing system of a language is transparent (Turkish, Spanish, Italian,...), opaque (Hebrew, and English) or semi-transparent (Farsi, Arabic,...), see the grain size theory in this regard

About the study from Cambridge, it never happened (i'ts just an email meme). But see what Matt Davies (actually from Cambridge) has to say about the postulates in the email, here: http://www.mrc-cbu.cam.ac.uk/people/matt.davis/Cmabrigde/

Grainger & Whitney published an informative commentary on the Cambridge whole word reading email in TICS, found here: http://www.cs.umd.edu/~shankar/cwhitney/Papers/TICS.pdf

Grainger & Whitney suggest that "open boigrams" are the units of encoding when reading Latin letters. But perhaps fast parallel processing of strings of single letters (where the letters are the units) is sufficient for fluent reading.

Congrat's!

verrry good question, but... not complete, because it's not clear about what 'reading': text or schemes.

it' used tio say that smallest quant is 'phrase is smallest containing sense' but for me "A sentence is a group of words that makes complete(!) sense" - the key expression here is 'complete sense' ... and it's matter of researcher's paradigm - for what purpose are you investigating and from what point of view and how far (what scope is) - 'reading' function for me is to get semantic sense.

some pop sources

http://www.brighthubeducation.com/help-with-writing/20570-what-makes-a-group-of-words-a-sentence/

http://cnx.org/content/m14812/latest/

http://www.brighthubeducation.com/help-with-writing/20570-what-makes-a-group-of-words-a-sentence/

do you know that book: " A functional English grammar. Margaret M. Bryant "

& related sources

http://en.wikipedia.org/wiki/Systemic_functional_grammar

http://en.wikipedia.org/wiki/Speech_acts

Thanks everybody for your comments. I'll try to comment back briefly. I agree with Paloma about parallel processing of several units, they should be of orthographic, phonologic and semantic nature. But MROM and other dual-route models don't work in parallel, they have lexical entries (it's too simple), but don't have morphemes. Grainger & Ziegler (2011) proposed a new version of this sort of models, with morpheme level and 'open bigrams' route (vote up to Randi). But morphemes and semantics levels in this model are mediated by a lexical level. It's not bad for ready, well-known words, but poses a problem for new words, like "Obamamania". Remember "Alice Through the Looking Glass": "Twas brillig, and the slithy toves // Did gyre and gimble in the wabe"... Feels like words, why? Because there are habitual and frequent letter combinations in proper orthographically plausible positions, these combinations are linked with some syllables which are linked with some meanings (not necessarily one to one; necessary to have links). I think, this is something that was meant by Osmar in the first of recent comments.

Maryam is right about difference between shallow and deep orthographies, but there should be some universal principles starting from the point that all orthographies are linked with phonology and semantics. Finnish, Chinese or Arabic persons all are humans and can successfully read with practically the same speed. Well, if not units, what are general principles of their reading?

Special thanks to Randi for the interesting link about "Cambridge research". I used Russian version of this text in my thesis, but didn't know the history of the question. About 'open bigrams' I have some doubts: they are units of computational encoding, but what psycholinguistic grounds do they have?

I have some more thoughts on these questions, if somebody's interested, but it'll be too much for one brief comment that I was going to write tonight.

Thanks, Andrius, for your vote. I'm not sure I quite understand you, I think, you're looking for a higher level of reading - comprehension. This level is independent of the mode of input, visual or audial. Levels of my interest are lower, starting ones. But they also have tight connection with meaning, with some units of sense. For example, prefixes and suffixes (of one, two or three letters) each have quite specific meaning.

As for functional grammar and language as a system - it is a traditional Soviet and Russian approach in linguistics, maybe, with some differences in details. Speech act theory is also studied by all undergraduate linguists. Again, I don't quite understand how these theories are related to my question.

Please take my word as some 'jazzing around' ;) cause I'm not in a classic linguistic's and not trying to pretend at all ;) but... I'm in business process and functional analysis, so from the view of systems theory I assume that it can't be gap between abstraction layers if we are climbing up'or'down system's hierarchy ladder's steps... I'd would like to point You to work of Bartini (see RU wiki) where he tries to show interconnections between physical units, so i think it must be some system's (functional) approach for the language unit's for reuse in IT (named entity recognition and related...) maybe some time later...

It's interesting: 'where comprehension start's ;) and how 'unit of reading' could relate to 'information unit' ?

Thanks, I'll see for Bartini.

Dmitry, the answer to your original question is "none of the above" or "all of the above" depending on how you view it!

Of course characters are to some extent phonetic, with younger languages with newer alphabets beings extremely phonetic, or at least phonemic (the precise sound can vary within a phoneme). Languages like English and French have become (or been deliberately) less phonemic. Russian orthography seems to have too many vowels, with its palalalized (ya, etc.) as well as unpalatalized (a, etc) sets, where it is not really clear that the y sound belongs with the vowel rather than the consonant. Chinese characters tend to have a meaning component and a phonemic or rhyming component, and of course when transcribing foreign words the latter is important.

The more phonemic a language is the more important the character can be in learning the language, and for spelling out and pronouncing unknown words - whilst with Chinese instinct can only give a rough indication of how an unknown character may be pronounced and what it means. Phonics is a method of teaching reading (in English) that is based on phonemic principles, but is controversial as it seems to slow down people's reading. Glenn Doman's teach your child to read system works on teaching them whole words, and letting them sort out the patterns unconsciously. Evelyn Wood's reading dynamics speed reading system, works on getting a reader (even an adult) to learn to focus on a whole paragraph (if the text is narrow and paragraphed enough) or early mid late chunks of the text using a serpentine handguided eye movement.

Phonics tends to keep reading rates down near speech rates, with 75 wpm being typical - an average typist can do as well. Word reading rates can get up to about 200 wpm (or max typing rates). Chunking at phrase to sentence level can get you to a comfortable 5000 wpm with full comprehension (and Evelyn Woods claimed improved comprehension), whilst chunking at paragraph to page level can get you to up to 40000 wpm (which is to say as limited mainly by how fast you can turn the pages (but Evelyn Woods use this scan speed for a prescan and postscan surrounding the 5000 wpm read for increased comprehension and retention).

There are of course disadvantages in reading for pleasure and finishing the book in half an hour (and I never used pre- or post-scans with fiction - who want's to know who done it before you've done your reading for enjoyment). And for technical material that you are actually learning from in a way that requires active integration in what you know, learning a new notation and new concepts before you can complete the article, a mixed mode approach is needed.

Reading machines used to be used to try to get school children to read faster, meaning to actually achieve 75 wpm plus. And in practice this tended to move them from a phoneme to a word oriented approach. The invention of the space is important for encouraging fast word oriented reading - without the space we are essentially forced to read phonemically. Evelyn Woods also experimented with ragged right and left lines, alternate indentation and outdentation, arguing that the extra cues helped you keep track of which line you were on and going to next, and that this helped with ( the line reading level of) speed reading.

Of course the active intervention of teaching Phonics, or Doman or Woods approaches, changes the way you read. The Australian Government (Federal Minister for Education, Nelson) opted to require early Phonics but encouraged an integrated approach where older children learn to rely less on sounding out words and develop fluency. However, Huey in the 1890s used eye tracking equipment to look at how people read without these interventions, and this showed actually phrasal level reading units rather than word or below, let alone sentence or above.

In particular, language have two type of word, functional (or closed class) words like 'the' or 'of' and contentive (or open class) words like nouns and verbs. The (inflectional or derivational) endings of words are also closed class and functional in nature. Closed vs open refers to the fact that a language is characterized by its functional or morphosyntactic elements and that changing these is not possible without changing the language; whilst the open class of content words changes on a daily basis as new words (and things and activities) are coined (or invented). Huey showed that the eye movements basically revolved around phrasal chunks that were mediated by functional words - it seems we pick up on the shape of the functional words subconsciously without them every being the object of foveal focus, and these then provide the framework or template for understanding the sentence - fitting the concepts associated with the word into a growing semantic conceptualization of the sentence and beyond.

The same sort of things happen in speech, with the functional words not only tending to be smaller (by Zipf's Law) but also pronounced faster, less stressed and in a reduced or clitical form - that is only vestigial elements may be present and they are attached to an associated word. We see that particularly with places where we put an apostrophe (consider it's and 'tis) but also there tends to be a drop of voicing (consider "a list of things I have to do" with voiced /v/ becoming unvoiced [f] as you add the functional/modal role of obligation as opposed to merely possessing the list). The prosody is also important to this, and this functional/prosodic structure of their mother tongue appears to be learned in the womb in the last trimester and already present at birth. Phonics and even word-level reading is arguably a backward step for language in view of this phrasal quality to speech. Of course in languages that have gender, the articles, and often adjectival contexts and declension, are presented with the nouns when teaching vocabulary, which is actually more consistent with how people use language whether reading or writing.

So the answer to your questions, from Huey 1998, is that phrases are the primary unit of reading, but that words and characters and morphemes (and given they are pronounceable as such, syllables) are important units too (but morphemes may actually be fused into a single syllable, and in faster reading or scanning even larger units at the length of clause, sentence or paragraph, can be the units.

David

That should be Huey 1898!

What came first? All the concepts and all theories or the acts of the mind? We really do have to start it all again following the neurosciences remarkable epistemology in order to test and prove all the theories of so diverse fields of linguistics knowledge.Maybe we can find some right and simple answers that may shape new paradigms! We cannot let it all apart as if old theories would had been written the same if the authors have had such opportunity.

I think the answer to this question depends very much on what "unit" you're referring. In the eye movement literature (K. Rayner and R. Kliegl as representatives), it is kind of a common consensus that word is the basic "perceptual encoding unit" in reading alphabetic scripts, such that readers target their eyes towards words (which makes intuitive sense given that words are marked by spaces). However, more recently, there are also evidence showing that morpheme properties also "affect" reading, at least during parafoveal preview (Drieghe et al., 2010; Angele & Rayner, 2012).

In Chinese, the situation is much more complicated. Given that word boundary is not explicit in Chinese and that Chinese readers disagree about word segmentation, it has been suggested that Chinese readers should encode each individual characters during reading (e.g., Chen et al., 2003: Developmental characteristics of eye movements in reading Chinese; Tsang & Chen, in press: Eye movement control in reading: Logographic Chinese versus alphabetic scripts). On the other hand, there are also recent evidence showing that explicit marking of word boundary helps Chinese readers to read faster, which appears to indicate word-based encoding (e.g. Blythe, 2012; Yan et al., 2012).

I guess the debate won't end easily but I do see a trend: On the English side, people are now moving from word to morpheme, but on the Chinese side, people are moving from character (morpheme) to word. Fun, right?

As a learner of Chinese, I feel a lot is lost by vocabulary being given in terms of words without explaining the characters (including both their morphemic and phonemic components) - it takes a lot more exposure to generalize synchronically, and the diachronic information can never be fully recovered. Similarly with learning English or other alphabetic languages, it is useful to explain some of the roots and affixes (particularly prefixes borrowed from Greek or Latin), although of course it is possible and usual to learn the language from examples without this additional information. And once the language has been bootstrapped to the point of self-learning even the learner of a foreign language can learn in this contextual way.

But another issue is the lack of a viable definition of word that works across languages, other than the backward one of saying it is those sequences of graphemes that are separated by spaces (or other punctuation). Why in English do we have "into" but "out of" ? why in German do we have "auschreiben" and "shreib aus" and call it one word (a separable verb) where in English we have "write out" and call it two words?

Diachronically, what are separate words in a pidgin, come together more as clitics in a creole and merge further as affixes as a language matures. But in the end the orthography only approximates roughly and inconsistent something that is a complex function of morphonology and prosody. Words are "meant" to be the smallest units you can say on their own, and normally are meant to be indivisible (broken by German separable verbs as well as im-#%&*-possible imprecations in English). We also have a special hyphenation convention in English for things that we can say and understand in speech in- or output but then have difficulty transcribing as indivisible words. One could just as easily say that the word "a" can't stand alone as way that the negatory prefix "a-" can't stand alone, they both can it is just that it is a- or untypical.

In the end, orthography is a set of prescriptive rules, deliberately stabilizing the written language with the advent of the printing press and the dictionary with the likes of Caxton and Webster. As far as I am concerned, there is really no such thing as a word in any meaningful crosslinguistic sense, and I also have reservations about the sentence in pre-literate speech. The phonological and prosodic 'words' have very little to do with the dictates of orthography and punctuation, which is one reason why I concur with Huey that they unit of reading is larger than both the conventional character and word.

dP

You all say very good things. Some of them I support in my parallel topic "Is lexicality a categorical or a gradual notion?". Here I'll try to describe my view on the title question.

Charles Perfetti and colleagues (2005; 2006) explicitly defined a word as a sum of three constituents: orthography, phonology and semantics. They integrate to constitute a word. They are sometimes plotted as three circles with interacting arrows. I offer to look at these circles as a bird from the sky see three neighboring towers. Let's look now not from up, but from side. All three constituents are hierarchies. Orthography starts from strokes, then letters (n-grams) or radicals in Chinese, orthographic words or Chinese characters and then higher to graphic means of marking dialogs, paragraphs, pages of text, etc. Phonology begins with phonemes, then syllables, phonetic words, syntagmas (prosodic constructions), utterance intonation and higher. Semantics consists of affixes, word roots, lexemes, propositions, frames and higher semantic constructions. Perception starts from the bottom levels, and rises through interaction of the three hierarchies. The unit of visual speech perception (reading) is an operation of closing the loop (circuit) between the three constituents. It's like drawing a triangle - it is not a triangle until you have three line segments closed together. At some level perception (activation) reaches orthographic word and is at or close to lexeme and phonetic word level. This degree of activation we usually consider a word.