Skip to main content
SpringerLink
Log in

Automaticity in fast lexical decision sequential effects: much like telling left from right

  • Original Article
  • Published:
Psychological Research Aims and scope Submit manuscript

Abstract

Successive lexical decisions have shown sequential effects where faster word responses and slower nonword responses follow the same versus different prior response. To date, explanations of these effects have been based on processes specific to discriminating words from nonwords. However, a more parsimonious explanation is possible, based on generic choice processes that apply even to left/right discriminations. Under conditions that promote automaticity, this explanation distinctly predicts equal facilitation by response repetition for words and nonwords. This hypothesis was here tested in an experiment involving 82 participants completing 850-trial blocks of lexical decision with a 100 ms response-stimulus interval—a much faster rate of choice succession than previously used—and including a factor of word/nonword discriminability so as to further test the applicability of choice-specific processes. Distinct from earlier findings, sequential effects were found to be identical in sign and substance for words and nonwords. This reliably occurred as facilitation by repetition across the decile distribution of response-times, across high and low levels of word/nonword discriminability, within each block of the run, and in interaction with higher-order sequential effects involving up to four prior trials. The main effect of facilitation by repetition at the second-order was particularly strong, being equal in effect-size to the interactive effect of the word/nonword factor and word/nonword discriminability (η 2 = 0.61). Hence, generic choice processes appeared to be sufficient to produce lexical decision sequential effects, independently of choice-specific processes. The findings particularly suggested a primary role for automatic response-facilitation, with accuracy-monitoring and expectancy contributing to higher-order effects. The further role of choice-specific processes in these and other findings, and the utility of lexical decision in studying generic choice processes, are discussed.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6

Similar content being viewed by others

Notes

  1. d * is standardised mean difference with routine correction; viz., g (1 − 3/k), where (per available data) \(g = \frac{{\bar{x}_{\text{A}} - \bar{x}_{\text{R}} }}{{{\text{SD}}_{\text{diff}} }} = \frac{{\bar{x}_{\text{A}} - \bar{x}_{\text{R}} }}{{\sqrt {2{\text{MSE}}} }} = \frac{T}{{\sqrt {2N} }}\) and k = 4 (n A + n R − 2) − 1 (Lakens, 2013, Eq. 6; Nakagawa & Cuthill, 2007, Eq. 14). Data from Lima and Huntsman (1997) were not available for this analysis.

  2. Indeed, all studies claiming word FR and nonword FA compiled their target series by dependent uniform sampling and/or run-length restriction, i.e., no more than 3 or 4 words/nonwords in succession. Compiling target series in these ways is not commended for studying sequential effects as it deviates the ratio of alternation to repetition trials from unity, and so potentially provides strategic information (Blais, 2008; Cantor, 1969, pp. 158–159; French & Perruchet, 2009; Schvaneveldt & Chase, 1969, p. 3, n. 4). Moreover, this deviation would, more often than not, be unequal between response alternatives, with the possibility that one decision becomes more strategically conditioned than the other.

  3. For example, if the model has a long-term memory for three items, and a short-term memory for two items, and the strengths (S) for match and mismatch across memories are (per Ratcliff & McKoon, 1995) 1 and 0.2, respectively, and a nonword has strength of 0.1, then, with weights (w) of 0.1 and 0.9 for prior (p) and target (t) items, respectively, the compound-cue \(\sum {S_{p}^{{w_{p} }} S_{t}^{{w_{t} }} }\), summing over all long-term representations, for wW trials is the sum of 1.1 × 0.2.9, 0.2.1 × 1.9 and 0.2.1 × 0.2.9 for match of the prior, target and unpresented items, respectively. This gives 1.29, and, similarly, 1.17 for nW. for drift by greater familiarity to the word decision; and 0.30 for nN and 0.34 for wN trials, for drift by lesser familiarity to the nonword decision. FR by familiarity is thus 0.12 for words, and 0.04 for nonwords. And, assuming that variability increases as drift-rate declines, the latter small effect is also less reliable.

  4. Sequential runs were highly correlated with other indices of randomness, e.g., Pearson rs ranged from 0.91 to 0.98 for Z 2 from the runs tests, and Chi squares from ratio and Good’s serial tests of first-order sequences.

  5. Counter-balancing this arrangement might have been more ideal, but in an earlier experiment with the same response-box where 40 participants were free to map hands per response, all but one participant chose the presently used order. Fixing the arrangement also helped with administering other tasks later in the session.

  6. Analysing ERs with inverse-normal transformation (as in Smith, Smith, Provost, & Heathcote, 2009) produced the same effects with equivalent effect-sizes.

  7. This is the recommended Type 8 in Hyndman and Fan (1996): Take m = (p + 1)/3, where p is a proportion of N observations, and then j = [N p  + m], and g = N p  + 3 − j, so that, for x values indexed from 0, the quantile q = (1 − g) × j − 1 + gx j . Boundary conditions apply. A Perl implementation is available on request.

  8. These pairwise differences are tabulated as Table 2 for RTs, and Table 3 for ERs, in Supplementary Materials.

  9. Exemplifying this point, a post hoc analysis of the role of word frequency (f) was made with respect to the MROM-based account, and reviewer suggestion. Classifying f as per Perea and Carreiras (2003), there was no f-modulation of nonword FR; some sign of more reliable low-to-low-f than high-to-low-f word FR, but only substantially in high NL blocks; and the most reliable word effects were for low-to-high-f word repetitions; there was no high-to-high-f word effect. Thus there was no f-modulation consistent with the choice-specific models. These results are not offered for generalisation. Further details are provided in Supplementary Materials.

References

Download references

Acknowledgments

We are grateful for the close consideration by two (anonymous) reviewers of earlier drafts of this paper. We also acknowledge with abiding thanks that Jürgen Keil and Walter Slaghuis (at UTAS) provided instructive and instrumental support in conducting the presently reported and related studies.

Author information

Authors and Affiliations

  1. Psychology, University of Tasmania, Churchill Avenue, Sandy Bay, TAS, 7005, Australia

    Roderick Garton & John A. Davidson

Authors
  1. Roderick Garton
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. John A. Davidson
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Roderick Garton.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary material 1 (DOCX 52 kb)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Garton, R., Davidson, J.A. Automaticity in fast lexical decision sequential effects: much like telling left from right. Psychological Research 80, 685–701 (2016). https://doi.org/10.1007/s00426-015-0671-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00426-015-0671-z

Keywords

Search

Navigation