Elsevier

Consciousness and Cognition

Volume 17, Issue 4, December 2008, Pages 1063-1081
Consciousness and Cognition

Fast Pairs: A visual word recognition paradigm for measuring entrenchment, top-down effects, and subjective phenomenology

https://doi.org/10.1016/j.concog.2008.09.004Get rights and content

Abstract

When word pairs having a familiar order are sequentially flashed on a computer in their non-familiar order, (code zip), observers have a strong phenomenology of seeing them in familiar order (zip code). Reversal errors remained frequent even when participants obtained perceptual experience of reverse-display items by beginning with a block of longer-duration trials. A forced-choice order-detection procedure reduced but did not eliminate reversal errors, showing that “fast pairs” is a robust perceptual illusion. Even adjective + noun pairs (green skirt) showed reversal errors, and reversal errors increased with the log frequency of the word pair, consistent with a strong role for statistical processing at the level of multi-word units.

Introduction

Everyday observation leads us to expect that the order in which we perceive events is the order in which they occur (Reeves & Sperling, 1986). Exceptions to this are phenomenologically interesting and also a scientific puzzle to be solved (Dennett, 1991).

In this paper we introduce a new paradigm for exploring discrepancies between perceived and actual order. We call our paradigm “Fast pairs.” Pairs of words are briefly flashed on a computer screen, one word immediately after the other, with the second word followed by a pattern mask. Observers are asked to report the words they saw, in the order in which they saw them. For exposure durations ranging from 30 to 90 ms, observers spontaneously reversed word pairs having a familiar order, such as fees legal and step next. Observers denied that they were purposely reporting words in their familiar order or that they were unsure of the order and thus simply guessed that the order was the familiar order. Instead they claimed they perceived the words in their familiar order.

It is well known that rapid visual displays lead to errors in reporting order (Lawrence, 1971, Norman, 1967, Reeves and Sperling, 1986). For example, in the “attention shift paradigm” (Reeves & Sperling, 1986), observers monitored letters displayed via RSVP. When a target was detected (for example, the letter “U”), observers were to report which number was current being displayed in a separate RSVP stream to the right to the right. When the numbers appearing after the target U were “6 2 1 7 3 4” observers frequently reported “7 3 1 4 2”, along with other less frequent incorrect orders. Reeves and Sperling’s explanation was that the attentional shift triggered by detection of the target letter ramps attention up to a maximum value, and the number that receives the peak of attention is then perceived as occurring first. The order of the numerals in visual short term memory is thus determined by the amount of attention allocated to each position in the numeral stream.

Another approach to studying errors in perceiving order are temporal-order-judgment (TOJ) tasks (e.g., Stolz, 1999), such as those used to investigate the “doctrine of prior entry” which states that attended stimuli are perceived prior to unattended stimuli (Boring, 1929, Nicole and Shore, 2007, Page and Norris, 1998, Shore et al., 2001). Despite this long-standing and on-going interest in TOJ tasks and their obvious relevance to questions of conscious perception (Libet, 1985), no researchers have investigated how familiar order influences perceptions of temporal order of words (for pictures in RSVP, see Intraub, 1989). Do highly familiar word pairs bias perception of temporal order more strongly than word pairs that are less familiar? To systematically manipulate familiarity, we studied two-word collocations. Collocations are frequently co-occurring word sequences that are generally recognized as familiar by fluent speakers of the language (Bybee et al., 2001, Harris, 1998, Jackendoff, 1995, Makkai, 1993, Wray, 2002). We compared high frequency collocations (e.g., thank you), low frequency collocations (e.g., gold medal), word pairs that were merely legal adjective + noun combinations (green skirt), and random combinations (pigs troops; see Table 1).

Following Langacker (1987), we will use the terms “entrenchment” and “entrenched” to refer to hypothetical mental processes of routinization and over-learning resulting from repeated exposure to meaningful stimuli. Compared to other patterns, entrenched patterns have robust long term memory representations and mental processing advantages, such as ease of recognition and recall. Entrenched patterns are frequently encountered, but frequency is a fact of the external world, while entrenchment describes the mental representations which may result from frequent exposure. Some highly frequent patterns may not become as entrenched as less frequent patterns if they lack conceptual coherence. For example, the phrase child abuse may be highly entrenched due to specificity of meaning and emotional impact, compared to highly frequent word pairs which don’t refer to a familiar concept, such as rather than or since you. In the current paper, participants’ familiarity ratings and text frequency counts are used as indications of entrenchment.

We first report a series of studies to document the effect of familiar order on word-pair recognition and the role of entrenchment on the temporal order illusion. We then discuss the two major questions that arise from these findings: Why do observers incorrectly perceive order, and how does their phenomenology compare to other order illusions? We will also describe the relevance of the fast-pair paradigm for Bayesian theories of word representation, and how it could be used for quantifying individual differences in susceptibility to top-down vs. bottom-up information sources.

Section snippets

Experiment 1: Measuring accuracy, reversals, and guessing

In the fast pairs paradigm, our (authors’ and lab assistants’) subjective phenomenology was that we were not uncertain about the temporal order: It was simply the order we experienced. It was disconcerting to learn that a stimulus had actually been fees legal when one had the experience of seeing legal fees. Study 1 verified our own experiences using controlled laboratory procedures. We examined whether participants were sometimes uncertain about the order by giving half of the participants the

Discussion

It is intriguing that display order did not influence identification accuracy. The stimulus about worry was recognized just as accurately as worry about. This could indicate that each word is identified independently of the word flashed before or after it. However, this appears to be ruled out by the finding that collocation status strongly influenced accuracy. Words that formed a low frequency collocation, even displayed in reversed order, were identified more accurately than words that were

Experiment 2: Varying exposure duration

We concluded from the low use of the “guess” option in Experiment 1 that participants had little subjective experience that they were reporting words in an incorrect order. Pilot studies revealed that warning participants that the words could be in reversed order did not reduce reversal errors. This suggests that participants may need actual perceptual experience of reversed-order words in order to adjust their default expectation of familiar order. To conclude that “fast pairs” is a robust

Discussion

The current results allow us to exclude the objection that participants reported canonical order because they did not fully believe that the items might be displayed in reversed order. In the 90 ms block, participants had improved perception of order, and could experience for themselves the fact that half of the items were frequently presented in non-familiar order. This knowledge reduced reversals, especially for the first, slowest block. For trials with shorter exposure, reversals were more

Experiment 3: Asking participants to choose between two orders

In the prior experiments, participants made reversal errors as part of trying to discern the identity of two briefly, sequentially displayed words. Reversal errors could be exacerbated by the cognitive effort involved in resolving word identity. That is, observers may expend maximal effort on word identification, with the result that order cues are ignored leading to the perception of familiar order. Reversal errors could also be influenced by the speech planning program which assembles the

Discussion

The forced-choice version of the fast pairs paradigm revealed that participants have difficulty determining display order even when they do not have to accurately discern the identity of the words under masking conditions or name them. As in the verbal report procedure (Experiments 1 and 2), the high frequency collocations also showed the largest difference in reversals between canonical and reversed display. Since participants did not verbally report their perceptions, reversal errors are not

Experiment 4: Forced choice with varying ISI

To further examine the robustness of the fast-pairs phenomenon, an experiment was conducted using the forced-choice methodology of Experiment 3, with the following modifications.

Discussion

Reversals occurred for both collocations and legal pairs, and the probability of reversals increased as a function of log Google frequency of the word pair, such that even word pairs labeled “random word pairs” demonstrated some reversals. Temporal order judgments were impaired for reverse-display word pairs using a newly created stimulus set employing a within-item design. This demonstrates that word frequencies (or other word-specific factors) are not responsible for the fast pairs illusion.

General discussion

Across the four experiments described in this article, familiar order strongly influenced the reported order of two sequentially, briefly displayed words. A number of findings were unprecedented in the literature:

  • Adjective + noun pairs displayed in unfamiliar order (skirt green) were spontaneously reported in their syntactically familiar order (green skirt).

  • Ability to correctly identify words was influenced by collocation status, but not by display order, and even adjective + noun pairs had a

Acknowledgments

The authors thank Jacqueline Liederman, Michael Mozer, Adam Reeves and Jennifer Stolz for helpful comments on a previous version of the manuscript, and Chloe Jordan for running participants in Experiment 4.

References (36)

  • J. Elman

    Finding structure in time

    Cognitive Science

    (1990)
  • D.A. Norman

    Temporal confusions and limited capacity processors

    Acta Psychologica

    (1967)
  • A.D. Barber et al.

    Cognitive control involved in overcoming prepotent response tendencies and switching between tasks

    Cerebral Cortex

    (2001)
  • E.G. Boring

    A history of experimental psychology

    (1929)
  • J. Bybee et al.

    Introduction

  • T.H. Carr

    Perceiving visual language

  • N. Chater et al.

    Probabilistic models of language processing and acquisition

    Trends in Cognitive Sciences

    (2006)
  • J.D. Cohen et al.

    PsyScope: An interactive graphic system for designing and controlling experiments in the psychology laboratory using Macintosh computers

    Behavior Research Methods, Instruments, and Computers

    (1993)
  • D.C. Dennett

    Consciousness explained

    (1991)
  • W.N. Francis et al.

    Frequency analysis of English usage: Lexicon and grammar

    (1982)
  • C.L. Harris

    Psycholinguistic studies of entrenchment

  • Harris, C. L. (1997). Distributed representations and mixed schemas. Presented at NIPS Workshop ‘Neural models of...
  • C.L. Harris

    Exploring the continuum of unit size in word identification

    (1996)
  • Hayes (2008). Query Google: A Java applet programmed by Tim Ma. Available from...
  • H. Intraub

    Illusory conjunctions of forms, objects and scenes during rapid serial visual search

    Journal of Experimental Psychology: Learning, Memory and Cognition

    (1989)
  • R. Jackendoff

    The boundaries of the lexicon

  • R.W. Langacker

    Foundations of cognitive grammar. Vol. I: Theoretical prerequisites

    (1987)
  • D.H. Lawrence

    Two studies of visual search for word targets with controlled rates of presentation

    Perception & Psychophysics

    (1971)
  • Cited by (0)

    The experiments were presented at the 69th Annual Meeting of the Eastern Psychological Association, Boston, MA, and the Eleventh International Conference on Cognitive and Neural Systems, Boston University, Boston, MA.

    View full text