Dehaene: Cerebral mechanisms of word masking and unconscious repetition priming (2001) Flashcards
What question is addressed by this research?
What brain activation is elicited during unconscious word processing (through masking) compared to conscious word processing?
What was the population and equipment used in the study?
The researchers used fMRI and EEG to measure the brain activity of 37 ‘right handed’ french students aged 19 to 34 years old.
The first experiment has 3M and 12F, the second experiment had 6M and 6F.
Give an overall picture of the (experimental) protocol used in the first experiment (e.g. what data was collected, e.g. neuroimaging method, behavioral task etc.).
The goal of experiment 1 was to image, within the circuit forword processing, the areas activated by masked words. fMRI and ERP’s were measured while participants viewed a random series of masking shapes and blank screens. This continuous visual stream was briefly interrupted by a short presentation of words that participants were asked to name in their head.
In one condition, the presence of blanks immediately surrounding the words made them consciously perceptible and reportable. In the other condition, the order of the masks and blanks was reversed so that the words were surrounded by masks that rendered them invisible
Two control situations were created in which the temporal context was identical but the words were omitted to isolate the activation caused solely by a masked or unmasked word.
Give an overall picture of the (experimental) protocol used in the second experiment (e.g. what data was collected, e.g. neuroimaging method, behavioral task etc.).
The goal of experiment 2 was to demonstrate that the masked words caused measurable repetition priming. Each trial contained a a short visual stream comprising a masked 29-ms prime word followed by a 500-ms target word. Because an overt naming task would have caused head motion, participants performed a semantic classification task by clicking with the left or righthand to indicate whether the target was natural or man-made respectively. On different trials, the prime and target could be the same word or a different word, and could appear in either lower or upper case.
The design allowed us to examine in which areas this repetition suppression was independent of case, indicating that the abstract identity of the letter string had been extracted. fMRI data was collected to observe repetition priming.
how could the researchers be sure that the words in the ‘unconscious’ condition were not consciously perceived? (4)
Several behavioural tests run immediately before and after brain imaging demonstrated that the masked words could not be detected, named or remembered.
Stimulus detection - did you detect the stimulus?
Word naming - participants were shown a continuous stream with 37 visible words, 37 masked words and 37 control blank trials appearing in random order at two-second intervals. Whenever they thought that a word had appeared, participants were asked to name it aloud or, if they felt unable to name it, merely say the word ‘vu’
Recognition memory - participants were shown each of the 37 visible words, the 37 masked words, and37 distractors that had never been presented before, in random order, for an unlimited time. They decided whether or not the word had been presented earlier.
Forced choice- 37 trials, a short stream comprising a single masked word was presented. Participants were told about the presence of a hidden word and were asked to select it among two choice words presented left and right of fixation.
Describe the fMRI methods used for data analysis (univariate contrasts, multivariate decoding etc.).
Functional connectivity analysis; For each participant the local maximum of left fusiform activity nearest to published coordinates was identified using fMRI. The residual of the above-described model provided an estimate of the fast variations of activity in this region, above and beyond the activation induced by the stimulation protocol. This residual was multiplied by the previous stimulus variables, and the resulting functions were entered as additional predictors in a new model of fMRI activity.
Two random-effect analyses were done. The first searched the whole brain for areas correlating with the left fusiform independently of the protocol . The other searched only the circuit activated by visible words for areas showing a significantly greater increase in correlation to visible words than to masked words, relative to their respective blank controls.
Describe the EEG methods used for data analysis (univariate contrasts, multivariate decoding etc.).
The researchers first identified three time windows in which groups of electrodes showed a significant difference between visible words and visible blanks. An analysis of variance on average voltage to the first stimulus of a trial tested for the main effect of visible stimulus (word or blank) and its interactions with hemisphere and time window when appropriate. We then applied the same analysis of variance to test for effects of the same sign and topography with unconscious words, using one-tailed tests and volt-ages averaged across the four words of a trial to increase statistical power.
Describe the main fMRI results of the first experiment (all experimental factors and dependent measures are important).
The fMRI data showed activation of a large fronto-parietal brain network, which disappears for unseen words.
There was also an increase in activation in the left visual word form area for visual words compared to invisible words however this activation did not disappear.
Describe the main EEG results of the first experiment (all experimental factors and dependent measures are important).
There was a standard sequence of events for word reading found: an initial occipital positivity (P1, 164ms), a left-lateralized occipitotemporal negativity (N1, 252ms), a short-lived central negativity (N400, 340 ms) and an extended central positivity (P3, 476 ms).
When contrasting masked words with masked blanks, an occipital P1 was also detectable, though it was smaller and slightly delayed.
The following N1 was similar in latency to the conscious N1 (peak at 252 ms), but that was strictly uni-lateral and restricted to left anterior temporal electrodes.
After this time point, masked and unmasked words differed radically. Neither an N400 nor a P3 was observed for masked words. Rather, the left N1 in response to masked words was pro-longed, and a second negativity restricted to left precentral electrodes was observed simultaneously with a focal central positivity.
Describe the main results of the second experiment (all experimental factors and dependent measures are important).
Reaction times were significantly shorter when the prime and target were the same word, independently of whether they appeared in the same case different-case priming.
Within the word-processing circuit, significant repetition suppression was observed in the left fusiform gyrus. At this site, priming was significant on same-case trials and on different-case trials. Case-independent priming was also found in the left precentral gyrus and in a symmetrical right precentral region.
Physical repetition priming restricted to same-case trials was observed in two right extrastriate regions.
Describe the conclusions that can be drawn from this article from the first experiment.
The design of experiment 1 allowed us to image the unconscious activity induced by isolated unseen words in the absence of any visible target and without any perceivable change in the ongoing stream. The results revealed a sequence of activations, with an early positive occipital waveform in ERPs plausibly corresponding to the extrastriate activation seen in fMRI, and two subsequent negative left-lateralized ERP components, which may correspond to the left fusiform and pre-central activations seen in fMRI. Thus, a complex processing stream that included high-level visual activity occurred in the absence of consciousness.
The conclusions that can be drawn from this article from the second experiment.
These results show that the repetition suppression phenomenon, which was previously obtained with consciously visible stimuli, can be replicated with unseen masked primes. As this phenomenon depends only on the identity of the masked prime, specific information about word identity must have been extracted and encoded unconsciously in the regions where repetition suppression was found.
Repetition suppression was case-specific in the right extrastriate cortex, and independent of case in the fusiform and precentral gyri. The hypothesis that the right extrastriate cortex is involved in feature-specific visual coding is supported by previous behavioral, PET and ERP studies.
Conversely, the case-independent priming effect indicates that the left fusiform region processes letter strings at a shape-invariant level.
What were the limitations of this study? (what could have been done better?)
Images of unconscious processing in experiment 1 were obtained by contrasting masked words with masked blanks. Thus, they might have reflected an undifferentiated burst of visual activity. (Thus exp. 2)
The study did not separate the contributions of letter-level, graphemic, phonological and semantic codes to the observed priming. By varying the level at which repetition suppression is occurring, priming offers a general tool to study the nature of the code in a given brain region. For instance, the hypothesis that the left precentral cortex is engaged in the phonological encoding of words could be tested by using homophonic primes (such as prime ‘one,’ target ‘won’).
Why is this study is relevant for the field of consciousness in general?
It demonstrates while we might not perceive a stimulus at a conscious level, our brain can process the stimulus to a certain degree.
