Elsevier

NeuroImage

Volume 34, Issue 4, 15 February 2007, Pages 1723-1732
NeuroImage

Spatial dynamics of masked picture repetition effects

https://doi.org/10.1016/j.neuroimage.2006.10.031Get rights and content

Abstract

The aim of this study was to further elucidate the mechanisms of early and automatic object processing using a masked picture priming paradigm with both identity and exemplar repetition in functional MRI (fMRI). Masked repetition priming has been commonly used with words to isolate automatic, rapidly occurring mechanisms involved in visual word recognition; however, studies using the technique of masked priming with rapid presentation of pictures have been limited. This study demonstrates how the masked priming technique can be used to study early, automatic processing of rapidly presented complex objects. Temporal–occipital regions previously found to be sensitive to repetition priming in both masked word and unmasked picture studies were found to show repetition suppression for the identity primes only. Most notably, when divided into anterior and posterior divisions, the fusiform gyrus showed anatomically specific repetition suppression only in the posterior portion. This finding is comparable to that found in a previous study of masked word priming, and the similarity may suggest an overlap in the early identification processes for visual word form and visual object processing in this region. Finally, masked repetition of different exemplar objects did not result in reliable neural effects, suggesting that the underlying mechanisms of the more semantic-based, different exemplar priming may occur later or require the intervention of conscious processes.

Introduction

The human visual system can rapidly process and identify objects on the order of a few hundred milliseconds. A complete understanding of the neural and perceptual processes that underlie human object recognition has proven difficult, in part because few techniques have sufficient temporal and spatial resolution to decompose the elemental components involved in recognition. Recently, Dehaene and colleagues (2004) have demonstrated that by combining fMRI recording with the masked repetition priming technique, it is possible to isolate the neuroanatomical substrates involved in the rapid bottom-up processing of visually presented words (Dehaene et al., 2004). In the current study, we extended the Dehaene et al. technique to the domain of object recognition, and report data from an experiment that is the first to combine fMRI recording and masked repetition priming with pictorial stimuli.

Various studies have established the neuroanatomical pathway involved in early and middle-level processing of visual stimuli. For this study we focused on the classic ventral system (Ungerleider and Mishkin, 1982) that cascades from V1 through V2 as well as V4 to the posterior TE and then to the entire TE (for a review see Tanaka, 1997). Yaginuma et al. (1993) have found that in monkeys, lesions to the TE, and along the pathway from V1 to TE, lead to disruptions in object recognition. These middle- and high-level processes involved in object recognition have also been studied in humans using fMRI in word and picture priming paradigms (e.g., Bar et al., 2001, Dehaene et al., 2001, Koutstaal et al., 2001).

Masked priming is characterized by a rapid presentation of stimuli. Typically the prime is presented very briefly and is immediately preceded and followed by a pattern mask and then a subsequent target stimulus (which is presented in the clear). This technique has been used in numerous behavioral studies to help decompose the time course of early word processing (Forster and Davis, 1984, Neely, 1991). The logic of this procedure is that the brief presentation of the prime event begins to activate representations in the visual system associated with the prime, but that the onset of the following mask stimulus interrupts processing of the prime prior to the level of conscious identification. By immediately following the mask with a target stimulus that is either the same or different than the prime, it is possible to probe for residual prime influence on the system by looking for subtle differences in target processing. By systematically manipulating the interval between the prime and target (what is called the SOA) and/or the duration of the prime itself, it is possible to infer different characteristics of early perceptual processing. The typical behavioral effect is that target processing is facilitated (e.g., RT for the target is faster) when the prime and target are the same stimulus or share certain features, compared to when they are different, “unrelated” stimuli (e.g., Bar and Biederman, 1998).

More recently, this paradigm has been adapted by Dehaene and colleagues for use with fMRI to examine the brain structures associated with early, automatic word processing. fMRI lacks temporal resolution due to the sluggish nature of the hemodynamic signal (Logothetis et al., 2001, Logothetis and Wandell, 2004), and therefore results are typically ambiguous with respect to whether a particular effect is due to feed-forward neural activity, recurrent feedback activity from higher brain areas, or both. As a possible solution to this, Dehaene et al. (2004) have argued that the mask priming technique, which effectively blocks feedback processing (Lamme and Roelfsema, 2000, Lamme et al., 2002), is a more precise tool for isolating brain structures involved in early recognition processes that are feed-forward (bottom-up) rather than feedback (top-down). In the current study we exploit this feature of masked priming to examine brain areas involved in early bottom-up aspects of object recognition.

Many fMRI studies have examined unmasked priming where both the prime and the target are available to conscious strategies (e.g., Buckner et al., 1998, Koutstaal et al., 2001, Vuilleumier et al., 2002); while relatively few studies have used the masked priming technique, particularly with visually presented objects. With relatively few masked picture priming studies, masked priming studies of words can nonetheless provide us with information about the similarities and differences of the processes involved in masked and unmasked priming.

In a recent language study, Dehaene et al. (2001) compared masked and unmasked repetition priming in words while participants made semantic decisions. Repetition suppression, similar to that seen in unmasked priming studies (decreases in the BOLD signal for repeated stimuli in comparison to unrepeated stimuli), was observed for masked primes as well. There was a greater reduction in activity for visible words when they were preceded by the same masked word as compared to words preceded by a different masked word. Differences in activation were undetectable in prefrontal and parietal areas, presumably reflecting the participants’ inability to report masked words. For repeated stimuli, where a masked prime was matched with an unmasked target, reductions were seen in the extrastriate cortex, fusiform and precentral gyri. The activation in the right extrastriate cortex was dependent on perceptual manipulations (i.e., lower case versus upper case). In the unmasked condition, the pattern of repetition-related changes was different, with an increase of activity evident in parietal, prefrontal and cingulate sites when compared to masked words. In the ERP portion of their experiment, only a P300 was observed for the unmasked condition and, therefore, they concluded that this may reflect the updating of consciousness (Dehaene et al., 2001).

Dehaene and colleagues (2004) contrasted these earlier results with a follow-up study where they examined masked word priming repetition effects for repeated words in the same retinal location and repeated words in different retinal locations. In the anterior and middle fusiform gyrus, repetition effects were seen for the same word, even when shifted by one letter position. Dehaene et al. (2004) suggested that, while the anterior and middle fusiform seem important in the coding of invariant features, the posterior fusiform effects support a more location-specific letter priming. They attributed the differential effect in the anterior–posterior subdivision of the fusiform to fusiform neurons that become progressively tuned from visual features in extrastriate cortex to larger units, such as graphemes or even words, as one moves anteriorly through fusiform gyrus. Since the right occipitotemporal cortex was only sensitive to repeated stimuli, the authors concluded that it is involved in processing low-level visual features, while a more abstract coding is performed by the left ventral occipitotemporal cortex where cross-case priming occurred.

In contrast to Dehaene’s work, Schnyer et al. (2002) observed repetition enhancement in their masked word repetition priming fMRI study. They hypothesized that if reductions in activation seen for unmasked repetition priming reflect facilitated perceptual processing, then these reductions should be observed in masked priming. However, if they are due to “post-identification attentional modulation,” then reductions may not be seen in masked priming (where there is no prime identification occurring). Following this line of reasoning, it might be suggested that differences in the pattern of increases and decreases in the fMRI signal associated with repetition may occur due to two processes: one earlier in the processing stream prior to identification and one later after identification has occurred.

While there has not been a report of masked priming with rapidly presented pictures using fMRI, a number of studies have examined the neural effects of picture repetition using supraliminal (unmasked) stimuli and lagged repetition (i.e., primes and targets separated by multiple intervening items). Using real and nonsense objects, Vuilleumier and colleagues (2002) examined repetition effects with objects of the same name, but with some difference in visual form. Decreases associated with repetition were observed in the bilateral inferior temporal gyri and posterior fusiform cortex, extending laterally into the right inferior occipital gyrus for both real and nonsense objects. The second presentation of real objects produced decreased activity in the bilateral fusiform and inferior frontal cortex. Repetition of different exemplars with the same name did not produce effects in the fusiform regions, from which Vuilleumier et al. (2002) concluded that the repetition effects observed for repeated real objects are the result of specific visual representations, not those reflected at a more abstract level such as lexical or semantic representations. Overall, the posterior fusiform was most affected (repetition-related decreases) by the repetition of real objects. The anterior fusiform was more engaged for meaningful stimuli since greater effects were observed for real objects than for nonsense objects. These effects were in the form of repetition suppression for repeated stimuli compared to the first exposure of the stimuli.

Two other studies found differences in the fusiform region when repetition of different exemplars was examined. Koutstaal et al. (2001) examined the neural correlates of visual object priming while participants made size judgments to pictures of objects that were exact repetitions or were different exemplars of items previously seen. They found repetition-related reductions across a wide range of brain regions in frontal and posterior areas. Of particular interest here was the modulation of the magnitude of repetition-related changes that occurred in the bilateral middle occipital, parahippocampal, and fusiform cortices when repetition spanned different visual exemplars. Occipitotemporal cortex (fusiform) showed laterality differences in sensitivity to identity/exemplar status, so that there was less repetition reduction for different exemplars in the right fusiform.

Similarly, Simons et al. (2003) found less repetition reduction in the right fusiform than in the left fusiform for repeated different exemplars. They argued that the left fusiform gyrus is more sensitive to lexical/semantic aspects of the stimuli while the right fusiform gyrus is more sensitive to perceptual form. Simons et al. (2003) also reported greater repetition reduction for different exemplars in the left occipitotemporal cortex. They interpreted this finding to indicate an increasing sensitivity to the complexity of object representations when moving from posterior to anterior locations in the left occipitotemporal cortex.

One possible reason for the discrepancy between Vuilleumier and colleagues findings with different exemplars and other studies involving different exemplar priming (Koutstaal et al., 2001, Simons et al., 2003) could be the difference in number of repetitions. While Simons et al. (2003) and Koutstaal et al. (2001) both have multiple repetitions of the stimuli, the Vuilleumier study used a single repetition.

The previously mentioned studies of supraliminal picture priming combined with the studies of masked word priming, suggest that there might be an overlap in the early visual areas utilized during word and picture recognition. However, several questions remain. First, the previous picture studies all used lagged repetition, which is known to modulate both perceptual and memory-based processes (Forster and Davis, 1984), thus leaving it unclear which of the observed effects are specific to object perception. Second, the use of supraliminal presentation in fMRI studies renders conclusions about the time course of repetition priming effects somewhat unclear (i.e., whether they are due to initial bottom-up processing, subsequent top-down processing or a combination of the two—see Dehaene et al., 2004). Interestingly, a recent event-related potential (ERP) study found evidence for early masked repetition effects to pictures of repeated objects (N/P190—Eddy et al., 2006), suggesting that masked repetition priming effects can be observed for single repetitions at short prime–target intervals. However, because of the relatively poor spatial resolution of ERPs, the authors were not able to locate the precise anatomical locus of their effects.

The current experiment aimed to examine the neuroanatomical loci of object repetition effects in an event-related fMRI design. Specifically, from this study we hoped to better understand which areas in the visual processing stream are affected by rapidly presented masked picture presentation and, more specifically, identify the underlying neural loci and mechanisms involved in identity repetition priming, as well as repetition of different exemplars. In summary, this experiment aims to address the following questions:

  • 1.

    Do masked repetitions of pictures produce suppression or enhancement in the BOLD signal?

  • 2.

    Is there evidence of an anterior–posterior dissociation within the fusiform gyrus in the response to masked picture repetitions?

  • 3.

    Is the functional neuroanatomy involved in masked picture priming a subset of the neuroanatomical loci involved in unmasked picture priming or are they separate substrates?

  • 4.

    Is there any evidence for masked repetition effects between different object exemplars or are such presumably semantically mediated effects dependent on identification of the prime (since the different exemplars share fewer physical similarities than the repeated exemplars, but have more semantic relatedness than the unrepeated condition)?

Section snippets

MRI participants

Seventeen Boston area undergraduates (8 females; mean age = 20.47 years) participated in this study for monetary compensation. They were all right handed with no history of neurological or psychiatric disorder, with normal or corrected-to-normal vision. Written informed consent was obtained from all participants. All procedures were approved by the Massachusetts General Hospital Human Studies Committee.

Stimuli

Real pictures of everyday objects were used in this experiment from Hemera photo-objects (//www.hemera.com

Results

Masked picture repetition effects were investigated using an event-related fMRI experiment. There were three critical conditions of interest: repetition where the masked prime and target were the same object, repetition with a different exemplar where the masked prime had the same name as the target, but had a different physical form, and unrepeated where the masked prime and target shared no physical or semantic similarities. Seventeen participants made button presses to non-critical food

Discussion

This experiment examined the effects of briefly presented masked pictures on the early and presumably bottom-up (automatic) processing of visual objects. Overall, the results revealed clear effects of masked picture priming within the ventral–temporal processing stream, up to and including the posterior fusiform gyrus. When comparing masked repeated pictures to masked unrepeated pictures, repetition suppression was observed in the inferior occipital gyrus, middle occipital gyrus, middle

Conclusions

The results of this study suggest that when pictures are presented rapidly and without conscious awareness, repetition suppression, rather than enhancement occurs, possibly reflecting facilitated perceptual processing (see Grill-Spector et al., 2006, Grill-Spector and Malach, 2001). In terms of visual object processing, the results also reveal some important findings with regards to similarities and differences in the way visual stimuli or stimuli that induce mental imagery (such as visual

Acknowledgments

This research was funded through grants HD025889 and HD043251 to Phillip J. Holcomb and K23MH64004 to David Schnyer. We thank Lindsay Nicholls for her assistance with the data collection.

References (38)

  • J.S. Simons et al.

    Neural mechanisms of visual object priming: evidence for perceptual and semantic distinctions in fusiform cortex

    NeuroImage

    (2003)
  • K. Tanaka

    Mechanisms of visual object recognition: monkey and human studies

    Curr. Opin. Neurobiol.

    (1997)
  • S.L. Thompson-Schill et al.

    A neural basis for category and modality specificity of semantic knowledge

    Neuropsychologia

    (1999)
  • T. Allison et al.

    Face recognition in human extrastriate cortex

    J. Neurophysiol.

    (1994)
  • M. Bar et al.

    Subliminal visual priming

    Psychol. Sci.

    (1998)
  • M. Bar et al.

    Cortical mechanisms of explicit visual object recognition

    Neuron

    (1998)
  • L. Cohen et al.

    Language-specific tuning of visual cortex? Functional properties of the visual word form area

    Brain

    (2002)
  • S. Dehaene et al.

    Cerebral mechanisms of word masking and unconscious repetition priming

    Nat. Neurosci.

    (2001)
  • S. Dehaene et al.

    Letter binding and invariant recognition of masked words: behavioral and neuroimaging evidence

    Psychol. Sci.

    (2004)
  • Cited by (17)

    • Semantic and perceptual priming activate partially overlapping brain networks as revealed by direct cortical recordings in humans

      2019, NeuroImage
      Citation Excerpt :

      However, given the labor-intensive process of matching the objects for the similarity, a number of studies employed an approach that differs from ours. These studies (scalp-EEG (Xu et al., 2012; Kiefer, 2001) and fMRI (Lucia et al., 2010; Schacter et al., 2004; Eddy et al., 2007)) used paradigms with repetition of identical or different exemplars from the same category. However, given that the different exemplars from the same category share visual features, which has been shown recently to be useful for discrimination between categories (Rupp et al., 2017), using this strategy, it is not possible to completely discern responses to semantic and perceptual priming.

    • A review and synthesis of the first 20years of PET and fMRI studies of heard speech, spoken language and reading

      2012, NeuroImage
      Citation Excerpt :

      It also explains why left occipitotemporal activation is sensitive to the left-right orientation of single letters and words but not to pictures (Dehaene et al., 2010a; Pegado et al., 2011) in terms of the learnt relationship between the visual form and higher-level language associations (which are orientation-specific for letters/words but not for objects). The early influence of language on visual word processing in the left ventral occipitotemporal cortex is consistent with observations that (a) the response to written words in a left-lateralized inferior frontal region (pars opercularis) peaks at the same time as that in the left ventral occipitotemporal cortex (Cornelissen et al., 2009); and (b) activation during picture naming or reading aloud is reduced when the target stimulus to be named is preceded by an unconscious masked prime that has the same name as the target but a different physical form (Eddy et al., 2007) as when a word is primed by a picture or a picture is primed by a word (Kherif et al., 2011). Finally, the degree to which the response in the ventral occipitotemporal cortex was left lateralized for words was found to correlate with the degree to which inferior frontal activation was left lateralized during word generation (Cai et al., 2010).

    • Invariance to rotation in depth measured by masked repetition priming is dependent on prime duration

      2011, Brain Research
      Citation Excerpt :

      This is because of the brief prime exposure and immediate presentation of the backward mask, which has been shown to cut-off recurrent processing (Lamme et al., 2002). Secondly, using the same masked repetition priming procedure in fMRI (although not with rotations in depth) Eddy et al. (2007) found no evidence for repetition suppression effects outside the fusiform gyrus. These findings do not eliminate the possibility that top–down feedback plays a role in object recognition, but do suggest that the repetition effects we observed with ERPs using masked priming are not likely produced by top–down facilitation.

    • Opposing patterns of neural priming in same-exemplar vs. different-exemplar repetition predict subsequent memory

      2011, NeuroImage
      Citation Excerpt :

      Repeated presentation of a stimulus often facilitates behavioral performance (e.g., decrease in response time) and decreases activation in task-specific brain regions (Buckner and Koutstaal, 1998; Grill-Spector et al., 1999; Henson, 2003; Schacter and Buckner, 1998; Wig et al., 2005; Wiggs and Martin, 1998). Recent studies have established that the magnitude of neural attenuation in the task-specific brain regions (e.g., inferior temporal, fusiform and inferior frontal cortices) depends on the perceptual similarity between the two stimulus presentations (e.g., Eddy et al., 2007; Koutstaal et al., 2001; Simons et al., 2003). Koutstaal et al. (2001) found larger attenuation in occipito-temporal and prefrontal cortices when repetitions involved perceptually identical stimuli (i.e., same-exemplar priming) than when they involved different exemplars of the same category (i.e., different-exemplar priming).

    View all citing articles on Scopus
    View full text