Gilbert and Sigman point out that response of neurons in the visual cortex are dependent upon the context in which an object appears… “a single oriented line segment will elicit a brisk response” … However, “when the line is embedded within a complex background of randomly oriented and positioned line elements, the neuron’s response is substantially inhibited”

– Strauss et al.

The parallel in reading faces is that if you want to know what emotion someone is feeling, you must watch the temporary changes in the face, because it is these rapid facial signals which give information about emotions.

– Ekman & Friesen

Neuroscience supports a context dependent, prediction based, “meaning construction view of the reading process” (Strauss et al., 2009). Brain doctors or neurologists, including Dr. Steven L. Strauss, MD have mounted a vigorous argument in support of the proposition that the brain in being efficient with limited resources, engages in prediction and selection of visual cues to make meaning through reading.

They posit the view that the way the brain deals with reading is no different from “the basic way the brain constructs meaning in all other contexts” – through verifying predictions. These other contexts include people and situations.

Their paper was an argument because they had to register and justify their disagreement with another dominant view on reading, inherent in the “phonological processing model”. This other view suggested reading is primarily a bottom up process, which entails reception of visual cues through the eyes, processed through the thalamus, in the lower regions of the brain before transmission to the “higher cortical regions” where meaning is made. The phonological processing model suggests words on a page must first be converted to sounds or “phonemes” and translated to oral language before meaning can be derived.

The main implication of the phonological processing model is that reading instruction should be heavily weighted towards phonics instruction and decoding at the level of text. This model also implies that every word in a text matters to equal degree, that only when every word on a page has been deliberately processed, does meaning materialize.

If this model was representative of the reading process, students would have to read a comprehension passage like they do an oral passage. Interestingly, it is actually not apparent students process and retain meaning from oral passages the way they do when engaging comprehension texts. Reading during an oral examination would be over in under five minutes. How many students recall details from an oral passage the way they can from a comprehension passage, when they reach home after the attempt?

When reading oral passages, students are arguably focused on pronunciation, expression and variation in tone. In other words, they read only to deliver an oral rendition of written text and theoretically, it is not inconceivable that someone could read with expression and variation in tone without deep understanding.

For example, consider the following sentences. “Joined at its inferior midline, the thalamus branches into a right and left portion. With the exception of olfactory sensation, every sensory modality travels through the thalamus on its way to the cortex”. If (this is most unlikely), these lines were part of an oral passage, all a student has to do is sound out the words, pause at the commas and the full stops.

To read well without sounding like one is reading the Romanised version of Inuktitut script, certain assumptions about meaning would have to be made even during oral examinations. These assumptions need not be rigorously tested for the purpose of reading aloud. A student who assumes ‘thalamus’ refers to a river, that the right and left portion’ to its tributaries, that ‘sensory modality’ is a fancy way of saying ‘boat’, that ‘olfactory sensation’ refers to a large ship and that ‘cortex’ refers to a port city, would really do no worse than someone who has the relevant schematic knowledge about human brain anatomy.

Strauss and his colleagues point to what they consider glaring omissions in the phonological processing model. For example, this view does not account for reading processes such as skimming and scanning, where the eyes move rapidly to get the gist of and to spot relevant information in passages.

They show that the cortex controls the thalamus and is not a “prisoner of the senses” and as evidence they cite that there are “ten times as many corticothalamic neurons as there are thalamocortical neurons”. Neurons which start at the cortex and travel to the thalamus are labelled corticothalamic. In their view, when facing a text, the brain makes predictions contingent on various cue types (semantics, syntax and schema) and instructs the thalamus on what stimulus to pick up from the text and let through to the cortex to confirm or reject a hypothesis about meaning in the text. This is how scanning happens.

To fortify this view, they refer to eye movement tracking and miscue studies. The former uses sophisticated technology to track how the eye moves during reading. The latter records how readers misread or omit words when they read aloud. Taken together, these studies show in what is claimed to be a conclusive way, that the cortex makes predictions about a text and the eye obeys its instructions to test those predictions.

In essence, they suggest students misread a text because of what they expect it to say rather than what it actually says. This evidences cortical control. This could mean that to read an oral passage without miscues, a student has to shut off thinking about the meaning of the passage. During preparation, he could make assumptions about meaning and decide the manner in which the rendition is to be delivered. Once he is reading aloud, he confines himself to merely performing as rehearsed.

Eye movement studies also show that readers skip “at least 20 to 30%” (Rayner,1997) of words during silent reading. Typically, function words such as determiners and connectors are skipped more than words bearing semantic content. Despite this, readers do form an accurate view on passages with the help of context and schema.

To be sure, Strauss and his colleagues are not suggesting that the reading process is entirely or by default a top down process, only that both top down and bottom up processes interact to render meaning in an efficient way.

This understanding is consistent with how intelligence analysts make sense of the world relevant to them. In Superforecasting, Tetlock and Gardner document how the Intelligence Advanced Research Projects Activity (IARPA) runs prediction tournaments to answer questions like “whether there would be a fatal confrontation between vessels in the East China Sea” (p.103) within a specified time period. Using techniques comprising a combination of top down and bottom up processes both entailing, scrupulously choosing between bits of data, forecasters managed to predict the answer correctly just before the closing date.

Paul Ekman writes about reading faces to make meaning of interactions. This is not the ancient art of physiognomy where a person’s facial features are taken to be indicative of his character and status. He refers to the human face as a “multisignal, multimessage system” which needs to be interpreted by differentiating and selecting relevant cues for accurate understanding. In certain interactions where what is heard triggers an emotional response, hearers may have a vested interest in veiling their true feelings. Indeed, wisdom enjoins us not to wear emotions on our sleeves. Facial expressions do mislead.

Children especially are particularly reliant on facial expressions of adults to interpret valence of events. If they are unable to gather sufficient information from facial cues, they could remain in a state of anxiety for a while. Here too, it might be useful for them to engage in both bottom up and top down processes. A facial expression is a stimulus capable of conveying meaning through a bottom up process. To understand it fully, situating it in its larger context would be useful. A way to engage in a top down process is to predict an emotional response based on a context and then to focus on facial regions where such responses are likely to show. A bottom-up process of analysing a facial expression could confirm anger in someone. Including the top down process in choosing an appropriate response would ascertain where that anger comes from or is directed at.

In sum, students need to be able to go deeper than the superficial, be it in their work, interactions or when deciding how to position themselves in relation to what might happen in the future.

To see thoroughly, they would need to go beyond seeing through their eyes.

The Brain Dojo