The role of orthographic neighbourhood effects in lateralized lexical decision: a replication study and meta-analysis

The effect of orthographic neighbourhood size (N) on lexical decision reaction time differs when words are presented in the left or right visual fields. Evidence suggests a facilitatory N effect (i.e., faster reaction times for words with larger neighbourhoods) in the left visual field. However, the N effect in the right visual field remains controversial: it may have a weaker facilitative role or it may even be inhibitory. In a pre-registered online experiment, we replicated the interaction between N and visual field and provided support for an inhibitory N effect in the right visual field. We subsequently conducted a pre-registered systematic review and meta-analysis to synthesise the available evidence and determine the direction of N effects across visual fields. Based on the evidence, it would seem the effect is inhibitory in the right visual field. Furthermore, the size of the N effect is considerably smaller in the right visual field. Both studies revealed considerable heterogeneity between participants and studies, and we consider the implications of this for future work.


Introduction
217 contrast, in the RVF high N targets were responded to 10.5 ms slower than low N targets (see 218 Figure 2A-C). This pattern of effects was near identical to that reported by Perea (Table S1).   Andrews (1997) concluded that facilitatory effects of N are restricted 295 to low-frequency items. Thus selecting a relatively high proportion of high-frequency items has 296 the potential to diminish N effects under central presentation. Indeed, when applying a median 297 split to zipf frequency, the N effect was more pronounced for low-frequency words (-13.7 ms) 298 than for high-frequency words (-1.0 ms).

299
In comparison to the published literature on lateralized N effects, the reaction times in the 300 current study were considerably longer; as illustrated in Figure 2B.

314
In sum, our replication study adds weight to the claim that effects of N differ across 315 visual fields. However, one question that remains is whether the N effect is reliably different 316 from zero in the RVF. As the current replication study was powered to detect the interaction 317 between N and visual fields, we decided to conduct a systematic review and meta-analysis to 318 address this quandary.

319
Study 2: A systematic review and meta-analysis 320 Despite the consensus that N effects differ across visual fields, the direction and 321 magnitude of the N effect appear far more consistent in the LVF than the RVF. While most 322 studies report a facilitatory effect in the LVF, the nature of the effect seems somewhat 323 controversial in the RVF. Therefore, we conducted a systematic review and meta-analysis to 324 synthesise the available data and provide the first statistical analysis examining whether the N 325 effect was reliably different from zero in the LVF and RVF.

345
The study selection process is illustrated in Figure 4. The abstracts of articles identified 346 through database searches were screened for duplicates. All duplicates were then removed. Six 347 reviewers then screened titles and abstracts of 232 unique articles using the metagear package 348 (version 0.6; Lajeunesse, 2020), which was also used to store and manage bibliographic data

386
As each experiment adopted a within-subjects design, it was necessary to obtain an 387 estimate of the correlation between reaction times for high-and low-N words in each visual field.
388 As none of the experiments included in the meta-analysis reported correlations between reaction 389 times for low-and high-N words across visual fields, we contacted the first, last, and 390 corresponding authors for each of the six articles requesting both the full and aggregate data. The 391 authors indicated that they no longer had access to the data or we received no reply within one 392 month. Thus, we could not obtain an estimate of the correlation between mean reaction times in 393 the high-and low-N condition in either visual field from the published literature. As such, we 394 decided to use our replication study to derive estimates of these correlations. From this dataset, 395 the correlation between reaction times in the low-and high-N condition was r= .90 in the LVF 396 and r= .79 in the RVF. It is important to note that these correlations are stronger in magnitude 397 than those typically observed in psychological science. This will of course have implications for 398 the outcome of our meta-analysis; as the correlation coefficient increases, the estimate of the 399 variance for each study decreases and precision increases. As such, we felt it necessary to repeat 400 the meta-analyse assuming a number of values for r and report these in the Supplemental 401 Materials.

402
To estimate the size of the N effect across visual fields, reaction times in the low N 403 condition were subtracted from scores in the high N condition. 410 where S 1 is the standard deviation of the high N condition, S 2 is the standard deviation of the low 411 N condition, and r is the correlation between the means in the high and low N conditions. A 412 summary of the effect sizes for the included experiments are shown in Table 3.

413
[Insert Table 3 here] 414 Publication bias 415 We assessed publication bias by presenting effect sizes against the inverse of their  Figure S2). However, the Egger's regression for the 423 RVF indicated no significant evidence for funnel plot asymmetry (t(9) = -2.06, p = .070).
424 Quality assessment

425
The quality of each study was assessed by two reviewers via the Appraisal Tool for    The results from the meta-analysis are shown in Table 4 and Figure 5. The estimated N 463 effect in the LVF was -31.8 ms, indicating that the N effect is facilitatory in the right hemisphere. 468 was larger for studies comparing the N effect in the RVF, indicating increased variation in 469 outcomes for the RVF relative to the LVF. This is an issue we return to in our meta-regression.

497
The source of between study heterogeneity may instead result from differences in 498 stimulus selection. For instance, the size of the N effect in each study will be influenced by 499 factors such as the presence of stimuli with body neighbours and/or a word's frequency of 500 occurrence in natural language. Body neighbours are orthographic neighbours that differ in 501 relation to the first letter position (e.g. house / mouse) (Andrews, 1997). Body neighbours 502 typically rhyme and have a facilitative effect on lexical decision times (Ziegler & Perry, 1998).
503 Thus, a study that has many words with body neighbours may encounter different N effects than 504 a study with fewer words with body neighbours. Similarly, as N effects appear stronger for low-505 frequency items (Andrews, 1997) it may be that studies which use a higher proportion of low-506 frequency items N effects of a different magnitude than a study using few low-frequency items.

524
These findings are compatible with the SERIOL model (Whitney, 2004). The SERIOL 525 model proposes that letter encoding for words presented in the LVF is inefficient or imprecise, 526 because there is a mismatch between the acuity gradient (based on the distance from fixation) 527 and the serial order of letters in the word (see Figure 1). As a result, top-down support from the 528 word level is required to disambiguate the word. Spreading activation among potential word 529 level targets results in more top-down support for words with larger neighbours: hence, N size

593
Finally, we would like to note that as Study 1 was conducted online, we could not control 594 for various environmental factors such as luminance, contrast, the participants' attention or 595 distance from the screen, and these may explain the variability in our data. It is reassuring 596 nonetheless that we managed to replicate the previous literature despite potential room for 597 increased noise. One way to add credibility to our novel findings would be to examine the 598 variability between participants with data collected under lab condition.

600
Two decades ago it was first reported that words with many orthographic neighbours 601 were processed differently in each visual field (Lavidor & Ellis, 2001). Since then, numerous 602 studies have investigated this interaction with many claiming that the N effect is faciliatory in the    Manuscript to be reviewed  Manuscript to be reviewed