Elsevier

Consciousness and Cognition

Volume 16, Issue 4, December 2007, Pages 942-953
Consciousness and Cognition

Unconscious semantic priming in the absence of partial awareness

https://doi.org/10.1016/j.concog.2006.07.003Get rights and content

Abstract

In a recent paper in Psychological Science, Kouider and Dupoux reported obtaining unconscious Stroop priming only when subjects had partial awareness of the masked distractor words (i.e., could consciously perceive subword features that enabled reconstruction of whole words). Kouider and Dupoux conjectured that semantic priming occurs only when such partial awareness is present. The present experiments tested this conjecture in an affective categorization priming task that differed from Kouider and Dupoux’s in using masked distractors that subjects had practiced earlier as visible words. Experiment 1 showed priming from practiced words when subjects had no partial awareness of those words. Experiment 2 showed that, in the absence of partial awareness, practiced words yielded priming but not-practiced words did not. Experiment 3 corroborated Experiment 1 and 2s results using a different test of partial awareness. These results suggest that unconscious processing (rather than partial awareness) of subword elements drives masked semantic priming by practiced words.

Introduction

An idea of long standing in the literature on attention is that for stimuli whose appearance is predicted by experience and expectation, unattended stimulus identification may require no more than processing at the level of individual features. In the domain of visual processing in which the present research takes place, such an account has been applied in the interpretation of priming effects from unattended distractors. For example, Stroop priming (the influence of a color-denoting distractor on color classification of a target stimulus) occurs even when the distractors consist of only the first few letters of a color word (Singer, Lappin, & Moore, 1975). Here, the context provides the expectation that a color word will appear, and distractors function as if they denote colors even though only partial processing is possible. In non-Stroop priming Broadbent and Gathercole (1990), conjectured that earlier findings that had been interpreted as showing semantic processing were in fact the result of processing at the feature level. Broadbent and Gathercole demonstrated that priming from unattended distractors (words presented peripherally) occurred only when the distractors had first appeared as centrally presented, attended targets; presumably, attended experience with the words as targets enabled unattended identification through processing of subword features.

More recently, the notion that context and experience enable unattended identification through processing of individual features has been revisited in priming studies that use masked, briefly flashed distractors that are consciously unidentifiable. In such studies a masked word distractor from one of two contrasting semantic categories (for example, pleasant-meaning versus unpleasant-meaning) precedes a nonmasked, visible target word from the same or contrasting category. The priming effect emerges as faster or more accurate target categorization on congruent trials (distractor and target from the same category) compared to incongruent trials.1 Such priming effects appear to be driven by part-word processing that enables distractor recognition only after earlier practice with the distractors as attended visible words. The evidence that such priming occurs at the subword level is of two kinds. First, studies that have directly compared priming from practiced and non-practiced distractors have applied an interpretation similar to that of Broadbent and Gathercole (1990). That is, they have interpreted a finding of unconscious priming from practiced distractors, but not from non-practiced distractors, as evidence that distractor processing is limited to individual features and includes little or no semantic analysis. Studies that have made this direct comparison with word distractors typically have obtained this pattern of findings (Abrams and Greenwald, 2000, Damian, 2001).2

The second form of evidence that unconscious priming is driven by processing at the level of subword elements is from studies that have manipulated distractor composition at the subword level. These studies have demonstrated priming effects that depend on the subword overlap between distractors and earlier-practiced words. For example, when nonwords (for example, anrm) are formed from parts of earlier-practiced words (angel and warm, categorized in practice as pleasant-meaning), those nonwords yield priming effects consistent with the valence of the practiced words from which they were formed (anrm functions as a pleasant-meaning distractor). When distractors of one valence (e.g., sad) are formed from parts of opposite-category practiced words (sun and glad), those distractors too function with the valence of the practiced words from which they were formed (sad functions as a pleasant-meaning distractor; Abrams & Greenwald, 2000). Additional evidence for a dominant influence of part-stimulus processing comes from a study of numbers that used a strategy similar to that in Abrams and Greenwald (Greenwald, Abrams, Naccache, & Dehaene, 2003). Collectively, these two forms of evidence—the difference between priming from practiced and non-practiced distractors, and priming that depends on subword overlap between distractors and practiced words—characterize unconscious priming as largely driven by processing that occurs at the level of subword features, and not by semantic processing.

A recent paper by Kouider and Dupoux (2004) casts doubt on a critical aspect of this notion that subword processing underlies unconscious priming. Kouider and Dupoux argue that, for masked primes under conditions of low visibility, priming is indeed driven by subword processing, but this subword processing is conscious rather than unconscious. Kouider and Dupoux introduce a contrast between partial and global awareness as an explanatory framework for interpreting findings of masked priming. Global awareness refers to the ability to consciously recognize whole-word meaning. Partial awareness refers to the ability to consciously identify only subword features. Partial awareness enables priming when there is a context that induces the expectation that certain words will appear, and subjects reconstruct expected whole-word meaning from consciously perceived subword elements. According to Kouider and Dupoux, priming can only occur when there is at least partial awareness. In studies that have reported unconscious priming, partial awareness must have been present, but was undetected because the typical measures of awareness are measures of global, not partial awareness.

Although Kouider and Dupoux used a Stroop task in the experiments that support their argument that priming requires at least partial awareness, they conjecture that the argument extends to non-Stroop forms of priming (“…we raise the conjecture that the only situations in which semantic priming is found are in cases of global awareness or partial awareness.”; Kouider & Dupoux, 2004, p. 81, in the context of a discussion of non-Stroop affective priming). This is the conjecture that we examined in the present research. One reason for suspecting that the partial awareness argument may not extend to all “situations in which semantic priming is found” is that in Kouider and Dupoux’s study the distractors had not been practiced, and as described above, priming occurs more reliably (at least in non-Stroop procedures) from practiced distractors. In Stroop priming, the distractors typically have not been practiced because the set of distractors (color words) is usually different from the set of targets (non-word color stimuli). In contrast, in non-Stroop tasks the distractors and targets are often drawn from the same set of words, allowing the continual practice throughout the task through the categorization of visible targets that then reappear as masked distractors.

Because earlier practice (in non-Stroop procedures) has been shown to so potently facilitate masked priming in comparison with priming without practice, it seems reasonable to consider the possibility that priming from practiced distractors may occur that does not require partial awareness. To test this, we first did an experiment that was similar to Kouider and Dupoux’s in its measure of partial awareness, but that used practiced distractors in a non-Stroop (affective categorization) task.

Section snippets

Experiment 1

The affective categorization procedure in Experiment 1 has been used widely in studies of unconscious semantic priming. The procedure, in overview, consists of a practice phase in which subjects repeatedly categorize visible targets as pleasant or unpleasant in meaning. A priming phase follows in which the just-practiced words reappear as both the visible targets and the masked distractors. The third phase of the task is typically a measure of, in Kouider and Dupoux’s terms, global awareness

Experiment 2

Experiment 1 obtained priming from practiced distractors when subjects did not have partial awareness of those distractors as defined by Kouider and Dupoux (2004). The purpose of Experiment 2 was to test the hypothesis, stated in the Introduction, that such priming without partial awareness occurs for practiced distractors, but not for unpracticed distractors like the ones used by Kouider and Dupoux. To carry out this test, Experiment 2 compared a condition in which distractors had been

Experiment 3

In preparing and testing Experiments 1 and 2, the authors found the measure of partial awareness somewhat more difficult to perform than expected. Interestingly, at some distractor durations that were tested informally, it was often easier to identify a word’s meaning than to say whether it was right-side up or upside down. Recognition of meaning seemed distinctly to precede identification of a word’s orientation. This observation at least suggests the possibility that the right-side-up

General discussion

The results of the present three experiments contradict Kouider and Dupoux’s (2004) conjecture that “the only situations in which semantic priming is found are in cases of global awareness or partial awareness.” In Experiments 1 and 2 semantic priming—that is, priming that depended on recognition of the distractor’s semantic category—was obtained under display conditions in which subjects performed at chance in the same measure of partial awareness that was used by Kouider and Dupoux. In

References (16)

  • K.A. Hutchison et al.

    Lexical and sub-lexical contributions to unconscious identity priming

    Consciousness and Cognition

    (2004)
  • R.L. Abrams et al.

    Parts outweigh the whole (word) in unconscious analysis of meaning

    Psychological Science

    (2000)
  • F.S. Bellezza et al.

    Words high and low in pleasantness as rated by male and female college students

    Behavior Research Methods, Instruments, and Computers

    (1986)
  • D.E. Broadbent et al.

    The processing of non-target words: semantic or not?

    Quarterly Journal of Experimental Psychology

    (1990)
  • J. Cohen

    Statistical power analysis for the behavioral sciences

    (1988)
  • M.F. Damian

    Congruity effects evoked by subliminally presented primes: Automaticity rather than semantic processing

    Journal of Experimental Psychology: Human Perception and Performance

    (2001)
  • A. De Clercq et al.

    A simple and sensitive method to measure timing accuracy

    Behavior Research Methods, Instruments, and Computers

    (2003)
  • S.C. Draine et al.

    Reliable unconscious semantic priming

    Journal of Experimental Psychology: General

    (1998)
There are more references available in the full text version of this article.

Cited by (0)

The authors thank the following Dickinson College undergraduates for their important contributions to the preparation of Experiment 3: Irena Ilieva, Alissa Niblock, Ritwik Niyogi, Carolyn Shainheit, and Anton Stoyanov.

View full text