The effect of a word


I have been avoiding, because of a lack of clarity, saying much about the relationship between language and consciousness. It is obviously important but hard to get a handle on. A recent article has prompted me to focus on this relationship. The article is by G. Lupyan and M. Spivey in PloS ONE, Making the Invisible Visible: Verbal by Not Visual Cues Enhance Visual Detection (here). Below is the abstract.

Can hearing a word change what one sees? Although visual sensitivity is known to be enhanced by attending to the location of the target, perceptual enhancements of following cues to the identity of an object have been difficult to find. Here, we show that perceptual sensitivity is enhanced by verbal, but not visual cues.

Participants completed an object detection task in which they made an object-presence or -absence decision to briefly-presented letters. Hearing the letter name prior to the detection task increased perceptual sensitivity. A visual cue in the form of a preview of the to-be-detected letter did not. Follow-up experiments found that the auditory cuing effect was specific to validly cued stimuli. The magnitude of the cuing effect positively correlated with an individual measure of vividness of mental imagery; introducing uncertainty into the position of the stimulus did not reduce the magnitude of the cuing effect, but eliminated the correlation with mental imagery.

Hearing a word made otherwise invisible objects visible. Interestingly, seeing a preview of the target stimulus did not similarly enhance detection of the target. These results are compatible with an account in which auditory verbal labels modulate lower-level visual processing. The findings show that a verbal cue in the form of hearing a word can influence even the most elementary visual processing and inform our understanding of how language affects perception.

To what extent can high-level cognitive expectation influence low-level sensory processing? Allocating visual attention to a location improves reaction times to probes appearing in that location. The spread of attention is also affected by specific objects: cuing an object speeds responses to a probe within the cued object’s boundaries.

We are dealing here with the edge between subliminal and conscious knowledge, where with a verbal cue the letter rises to conscious awareness but without the verbal cue it is not consciously seen. The results say a lot about perception, language and consciousness.

The results give conformation to the idea that there is top-down influence on very basic and early sensory perception.

The simple detection task is compatible with one of two broad conclusions: a) visual detection processes in visual cortex are influenced by auditory linguistic signals, or b) the process of detecting visual signals includes non-visual areas of cortex which are richly influenced by auditory linguistic signals. Either conclusion requires rejecting the assumption that “simple” visual tasks such as object detection depend only on the visual characteristics of a stimulus. … The present findings appear to conform to … requirements for … cognitive penetrability of early vision because information from outside the visual system (the linguistic label) is affecting visual sensitivity… We conclude based on the present findings that auditory verbal cues actually alter perceptual processing of the named objects rather than alter a higher level decision process.

What happens when an object that is being perceived has been given a name?

One way to understand our results is by conceiving of verbal labels as providing modulatory feedback to the visual system (The Label Feedback Hypothesis). Attention (one form of top-down control) has been shown to affect response properties of neurons in the very first visual area receiving top-down projections—the lateral geniculate nucleus (thalamus area)—and there is a large literature on effects of context, task-demands, and expectations on neural responses in primary visual cortex. The present results offer evidence that verbal labels, by virtue of their pre-existing association with visual stimuli, modulate visual processing by providing a “head-start” to the visual system, facilitating the processing of stimuli associated with the label. This type of continuous interaction between top-down and bottom-up processes is consistent with a number of theoretical frameworks

Currently ongoing experiments indicate that similar results can be obtained for pictures of everyday objects and animals: hearing common nouns can facilitate the detection of pictures from the named category… (Other results) suggest that the format of the cue, in addition to its modality, is important: verbal auditory cues (e.g., “cow”) facilitated visual identification and discrimination more than nonverbal auditory cues (e.g., the sound of a cow mooing”).

There is now accumulating evidence that higher level semantic information can influence visual perception in some surprising ways. For instance, auditory processing of verbs associated with particular directions of motion (e.g., fly, bomb) interferes with visual discrimination tasks along the vertical axis and increases sensitivity to the congruent motion direction in random-dot kinematograms. Moreover, linguistic input can guide visual search in an incremental and automatic fashion. Ascribing meaning to unfamiliar shapes using verbal labels improves the efficiency of visual search for these shapes. In fact, simply hearing a word that labels the target improves the speed and efficiency of search (compared to not hearing the label, but still knowing the target’s identity). For instance, when searching for the number 2 among 5′s, participants are faster to find the target when they actually hear “find the two” immediately prior to the search trial – even when they know that the 2 is the target because is has been so for the entire block of trials.

Words and their meaning have a great influence on the focus of attention, on the content of consciousness and on the details of perceptive processing. This is in keeping with Bolles’ model in the Babel’s Dawn blog (here) of speech being about joint attention, with words being the way to point attention to a particular topic.