Verbal Working Memory and On-Line Sentence Comprehension: An Evidence from The Visual-World Paradigm and Listening
2023-21-09
Methods
Participants
We recruited 55 participants (x males, Mage = , SD = , range = ). All participants were native speakers of Turkish. Participants reported normal or corrected-to-normal vision, no speech or hearing difficulties and no history of any neurological disorder. All participants were fully informed about the details of the experimental procedure and gave written consent. Post-experiment debriefing revealed that all participants were naïve to the purpose of the experiment.
Results
As dwell time includes saccadic movements in addition to fixations, we preferred fixation percentage over dwell time percentage as a more refined indicator of online sentence processing in visual world paradigm. Accordingly, we expected that participants would fixate on the relevant quadrant to derive support from the environment (see Kumcu & Thompson, 2020).
Dwell time percentage on the target grid
Random effects
Both random effects (i.e., subject and item) improve the null target model, thus they were added as random effects in the following target models.
## Data: target
## Models:
## re1: dtp ~ structure + distance + (1 | subject)
## re0: dtp ~ structure + distance + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## re1 5 -4687.6 -4652.2 2348.8 -4697.6
## re0 6 -6051.2 -6008.7 3031.6 -6063.2 1365.6 1 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: target
## Models:
## re2: dtp ~ structure + distance + (1 | item)
## re0: dtp ~ structure + distance + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## re2 5 -4341.1 -4305.7 2175.5 -4351.1
## re0 6 -6051.2 -6008.7 3031.6 -6063.2 1712.1 1 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1The effect of condition
The target model with condition as the fixed effect shows that condition (combination of structure and distance) significantly improves the null target model.
## Data: target
## Models:
## c0: dtp ~ (1 | subject) + (1 | item)
## c1: dtp ~ condition + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## c0 4 -6049.5 -6021.1 3028.7 -6057.5
## c1 7 -6051.3 -6001.8 3032.7 -6065.3 7.8689 3 0.0488 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Linear mixed model fit by maximum likelihood . t-tests use Satterthwaite's
## method [lmerModLmerTest]
## Formula: dtp ~ condition + (1 | subject) + (1 | item)
## Data: target
##
## AIC BIC logLik deviance df.resid
## -6051.3 -6001.8 3032.7 -6065.3 8790
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.4281 -0.6751 -0.1162 0.5708 4.1250
##
## Random effects:
## Groups Name Variance Std.Dev.
## item (Intercept) 0.006083 0.07799
## subject (Intercept) 0.006701 0.08186
## Residual 0.027403 0.16554
## Number of obs: 8797, groups: item, 160; subject, 55
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 0.260067 0.016922 180.289861 15.369 <2e-16 ***
## conditionlinear_short -0.046290 0.018140 159.443183 -2.552 0.0117 *
## conditionnested_long -0.003997 0.018140 159.443183 -0.220 0.8259
## conditionnested_short -0.012887 0.018140 159.443183 -0.710 0.4785
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) cndtnl_ cndtnnstd_l
## cndtnlnr_sh -0.536
## cndtnnstd_l -0.536 0.500
## cndtnnstd_s -0.536 0.500 0.500The effect of structure and distance
We therefore examined the effect of structure and distance separately. Individual models show that structure (i.e. linear and nested) is not a significant factor. However, distance is. In general, there are fewer looks in the target visual when processing short sentences. As the model with combination as a fixed effect suggested, the interaction of structure and distance has no significant effect.
## Data: target
## Models:
## l0: dtp ~ (1 | subject) + (1 | item)
## l1: dtp ~ structure + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## l0 4 -6049.5 -6021.1 3028.7 -6057.5
## l1 5 -6048.7 -6013.3 3029.4 -6058.7 1.2557 1 0.2625## Data: target
## Models:
## l0: dtp ~ (1 | subject) + (1 | item)
## l2: dtp ~ distance + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## l0 4 -6049.5 -6021.1 3028.7 -6057.5
## l2 5 -6051.9 -6016.5 3031.0 -6061.9 4.466 1 0.03458 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Linear mixed model fit by maximum likelihood . t-tests use Satterthwaite's
## method [lmerModLmerTest]
## Formula: dtp ~ distance + (1 | subject) + (1 | item)
## Data: target
##
## AIC BIC logLik deviance df.resid
## -6051.9 -6016.5 3031.0 -6061.9 8792
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.4379 -0.6751 -0.1160 0.5695 4.1235
##
## Random effects:
## Groups Name Variance Std.Dev.
## item (Intercept) 0.006225 0.07890
## subject (Intercept) 0.006701 0.08186
## Residual 0.027403 0.16554
## Number of obs: 8797, groups: item, 160; subject, 55
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 0.25807 0.01435 128.85287 17.986 <2e-16 ***
## distanceshort -0.02759 0.01296 159.45325 -2.128 0.0349 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## distancshrt -0.452## Data: target
## Models:
## l3: dtp ~ structure + distance + (1 | subject) + (1 | item)
## l0: dtp ~ structure * distance + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## l3 6 -6051.2 -6008.7 3031.6 -6063.2
## l0 7 -6051.3 -6001.8 3032.7 -6065.3 2.1114 1 0.1462##
## Descriptive statistics by group
## group: linear
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4399 0.24 0.2 0.18 0.22 0.21 0 1 1 0.88 0.32 0
## ------------------------------------------------------------
## group: nested
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4398 0.25 0.2 0.2 0.23 0.21 0 0.97 0.97 0.75 -0.08 0##
## Descriptive statistics by group
## group: long
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4399 0.26 0.2 0.22 0.24 0.23 0 1 1 0.74 0.09 0
## ------------------------------------------------------------
## group: short
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4398 0.23 0.2 0.17 0.21 0.19 0 0.97 0.97 0.89 0.16 0First run fixation duration on the target visual
As an addition, we also investigated first and last runs into the grids. There is no significant effect of distance or target on the looks in the target visual during the first run.
## Data: target
## Models:
## f0: firstfixdur ~ (1 | subject) + (1 | item)
## f1: firstfixdur ~ structure + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## f0 4 -16827 -16799 8417.4 -16835
## f1 5 -16827 -16792 8418.3 -16837 1.7348 1 0.1878## Data: target
## Models:
## f0: firstfixdur ~ (1 | subject) + (1 | item)
## f2: firstfixdur ~ distance + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## f0 4 -16827 -16799 8417.4 -16835
## f2 5 -16825 -16790 8417.5 -16835 0.1302 1 0.7183Last run fixation duration on the target visual
However, as the general pattern has shown, during the processing of short sentences in the last run of looking behaviour, there are shorter fixations in the target visual when processing shorter sentences as compared to longer sentences.Structure did not have an effect.
## Data: target
## Models:
## z0: lastfixdur ~ (1 | subject) + (1 | item)
## z1: lastfixdur ~ structure + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## z0 4 -13106 -13078 6557.0 -13114
## z1 5 -13104 -13069 6557.2 -13114 0.5629 1 0.4531## Data: target
## Models:
## z0: lastfixdur ~ (1 | subject) + (1 | item)
## z2: lastfixdur ~ distance + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## z0 4 -13106 -13078 6557.0 -13114
## z2 5 -13109 -13074 6559.5 -13119 5.041 1 0.02475 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Linear mixed model fit by maximum likelihood . t-tests use Satterthwaite's
## method [lmerModLmerTest]
## Formula: lastfixdur ~ distance + (1 | subject) + (1 | item)
## Data: target
##
## AIC BIC logLik deviance df.resid
## -13109.0 -13073.9 6559.5 -13119.0 8192
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.4659 -0.5697 -0.1980 0.3654 7.7263
##
## Random effects:
## Groups Name Variance Std.Dev.
## item (Intercept) 0.001058 0.03253
## subject (Intercept) 0.002783 0.05275
## Residual 0.011142 0.10555
## Number of obs: 8197, groups: item, 160; subject, 55
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 0.143660 0.008157 86.676533 17.612 <2e-16 ***
## distanceshort -0.012786 0.005650 159.282050 -2.263 0.025 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## distancshrt -0.346##
## Descriptive statistics by group
## group: linear
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4067 0.13 0.12 0.09 0.11 0.08 0.02 1 0.98 2.46 8.77 0
## ------------------------------------------------------------
## group: nested
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4130 0.14 0.12 0.1 0.12 0.08 0.02 0.96 0.94 2.21 7.33 0##
## Descriptive statistics by group
## group: long
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4143 0.14 0.13 0.1 0.12 0.08 0.02 1 0.98 2.34 8.07 0
## ------------------------------------------------------------
## group: short
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4054 0.13 0.12 0.09 0.11 0.07 0.02 0.96 0.94 2.29 7.66 0Working memory and target grid
There is no correlation between the working memory indicators and the percentage of time spent looking at the target visual.
##
## Pearson's product-moment correlation
##
## data: target_wm_all$dtp and target_wm_all$acc
## t = 0.7893, df = 54, p-value = 0.4334
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1606143 0.3595999
## sample estimates:
## cor
## 0.1067957##
## Pearson's product-moment correlation
##
## data: target_wm_all$dtp and target_wm_all$rt
## t = -0.26915, df = 54, p-value = 0.7888
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.2966476 0.2284977
## sample estimates:
## cor
## -0.03660157##
## Pearson's product-moment correlation
##
## data: target_wm_all$dtp and target_wm_all$mspan
## t = 0.75978, df = 54, p-value = 0.4507
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1645035 0.3561168
## sample estimates:
## cor
## 0.1028449##
## Pearson's product-moment correlation
##
## data: target_wm_all$dtp and target_wm_all$time
## t = 0.36459, df = 54, p-value = 0.7168
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.2161626 0.3084363
## sample estimates:
## cor
## 0.04955404##
## Pearson's product-moment correlation
##
## data: target_wm_long$dtp and target_wm_long$acc
## t = 1.0099, df = 53, p-value = 0.3171
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1327267 0.3885421
## sample estimates:
## cor
## 0.137409##
## Pearson's product-moment correlation
##
## data: target_wm_long$dtp and target_wm_long$rt
## t = -0.24728, df = 53, p-value = 0.8056
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.2965730 0.2334529
## sample estimates:
## cor
## -0.03394671##
## Pearson's product-moment correlation
##
## data: target_wm_long$dtp and target_wm_long$mspan
## t = 0.71728, df = 53, p-value = 0.4763
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.171712 0.354137
## sample estimates:
## cor
## 0.0980519##
## Pearson's product-moment correlation
##
## data: target_wm_long$dtp and target_wm_long$time
## t = 0.37983, df = 53, p-value = 0.7056
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.2161823 0.3130728
## sample estimates:
## cor
## 0.05210314##
## Pearson's product-moment correlation
##
## data: target_wm_short$dtp and target_wm_short$acc
## t = 0.49641, df = 54, p-value = 0.6216
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1990276 0.3245493
## sample estimates:
## cor
## 0.06739963##
## Pearson's product-moment correlation
##
## data: target_wm_short$dtp and target_wm_short$rt
## t = -0.27405, df = 54, p-value = 0.7851
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.2972559 0.2278653
## sample estimates:
## cor
## -0.03726782##
## Pearson's product-moment correlation
##
## data: target_wm_short$dtp and target_wm_short$mspan
## t = 0.75249, df = 54, p-value = 0.455
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1654633 0.3552549
## sample estimates:
## cor
## 0.1018686##
## Pearson's product-moment correlation
##
## data: target_wm_short$dtp and target_wm_short$time
## t = 0.32255, df = 54, p-value = 0.7483
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.2216042 0.3032556
## sample estimates:
## cor
## 0.04385111Dwell time percentage on the competitor grid
Random effects
Both random effects (i.e., subject and item) improve the null competitor model. However, the competitor models did not converge with item as random effect, thus they were fit with subjects only.
## Data: competitor
## Models:
## re1: dtp ~ structure + distance + (1 | subject)
## re0: dtp ~ structure + distance + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## re1 5 -4487.9 -4452.5 2248.9 -4497.9
## re0 6 -5034.2 -4991.7 2523.1 -5046.2 548.35 1 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: competitor
## Models:
## re2: dtp ~ structure + distance + (1 | item)
## re0: dtp ~ structure + distance + (1 | subject) + (1 | item)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## re2 5 -1381.4 -1346.0 695.68 -1391.4
## re0 6 -5034.2 -4991.7 2523.11 -5046.2 3654.9 1 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1The effect of condition
The competitor model with condition as the fixed effect shows that condition (combination of structure and distance) significantly improves the null competitor model.
## Data: competitor
## Models:
## c0: dtp ~ (1 | subject)
## c1: dtp ~ condition + (1 | subject)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## c0 3 -4459.5 -4438.3 2232.8 -4465.5
## c1 6 -4485.9 -4443.4 2248.9 -4497.9 32.36 3 4.395e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Linear mixed model fit by maximum likelihood . t-tests use Satterthwaite's
## method [lmerModLmerTest]
## Formula: dtp ~ condition + (1 | subject)
## Data: competitor
##
## AIC BIC logLik deviance df.resid
## -4485.9 -4443.4 2248.9 -4497.9 8791
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.0236 -0.5979 0.0113 0.6637 4.7640
##
## Random effects:
## Groups Name Variance Std.Dev.
## subject (Intercept) 0.01736 0.1318
## Residual 0.03416 0.1848
## Number of obs: 8797, groups: subject, 55
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 3.894e-01 1.820e-02 5.908e+01 21.397 < 2e-16 ***
## conditionlinear_short 2.179e-02 5.573e-03 8.742e+03 3.910 9.32e-05 ***
## conditionnested_long -3.698e-03 5.573e-03 8.742e+03 -0.664 0.506953
## conditionnested_short 1.890e-02 5.573e-03 8.742e+03 3.391 0.000699 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) cndtnl_ cndtnnstd_l
## cndtnlnr_sh -0.153
## cndtnnstd_l -0.153 0.500
## cndtnnstd_s -0.153 0.500 0.500The effect of structure and distance
We therefore examined the effect of structure and distance separately. Individual models show that structure (i.e. linear and nested) is not a significant factor. However, distance is. In general, there are more looks on the competitor visual when processing short sentences as compared to long sentences. As the model with combination as a fixed effect suggested, the interaction of structure and distance has no significant effect.
## Data: competitor
## Models:
## l0: dtp ~ (1 | subject)
## l1: dtp ~ structure + (1 | subject)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## l0 3 -4459.5 -4438.3 2232.8 -4465.5
## l1 4 -4458.2 -4429.9 2233.1 -4466.2 0.6949 1 0.4045## Data: competitor
## Models:
## l0: dtp ~ (1 | subject)
## l2: dtp ~ distance + (1 | subject)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## l0 3 -4459.5 -4438.3 2232.8 -4465.5
## l2 4 -4489.2 -4460.8 2248.6 -4497.2 31.651 1 1.845e-08 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Linear mixed model fit by maximum likelihood . t-tests use Satterthwaite's
## method [lmerModLmerTest]
## Formula: dtp ~ distance + (1 | subject)
## Data: competitor
##
## AIC BIC logLik deviance df.resid
## -4489.2 -4460.8 2248.6 -4497.2 8793
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.0335 -0.6014 0.0115 0.6641 4.7716
##
## Random effects:
## Groups Name Variance Std.Dev.
## subject (Intercept) 0.01736 0.1318
## Residual 0.03416 0.1848
## Number of obs: 8797, groups: subject, 55
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 3.876e-01 1.799e-02 5.634e+01 21.550 < 2e-16 ***
## distanceshort 2.219e-02 3.941e-03 8.742e+03 5.631 1.85e-08 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## distancshrt -0.110## Data: competitor
## Models:
## l3: dtp ~ structure + distance + (1 | subject)
## l0: dtp ~ structure * distance + (1 | subject)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## l3 5 -4487.9 -4452.5 2248.9 -4497.9
## l0 6 -4485.9 -4443.4 2248.9 -4497.9 0.0105 1 0.9182##
## Descriptive statistics by group
## group: linear
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4399 0.4 0.23 0.37 0.39 0.26 0 1 1 0.25 -0.8 0
## ------------------------------------------------------------
## group: nested
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4398 0.4 0.23 0.38 0.39 0.27 0 1 1 0.2 -0.83 0##
## Descriptive statistics by group
## group: long
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4399 0.39 0.22 0.36 0.38 0.25 0 1 1 0.29 -0.7 0
## ------------------------------------------------------------
## group: short
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4398 0.41 0.23 0.39 0.41 0.28 0 1 1 0.15 -0.91 0First run fixation duration on the competitor visual
As an addition, we also investigated first and last runs into the grids. Similar to the general viewing pattern, there is no significant effect of structure on the the looks in the competitor visual during the first run. However, as in the general viewing pattern, fixations are longer on the competitor visual when processing short sentences as compared to long sentences.
## Data: competitor
## Models:
## f0: firstfixdur ~ (1 | subject)
## f1: firstfixdur ~ structure + (1 | subject)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## f0 3 -13758 -13737 6882.0 -13764
## f1 4 -13758 -13730 6883.1 -13766 2.28 1 0.1311## Data: competitor
## Models:
## f0: firstfixdur ~ (1 | subject)
## f2: firstfixdur ~ distance + (1 | subject)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## f0 3 -13758 -13737 6882.0 -13764
## f2 4 -13762 -13733 6884.8 -13770 5.5669 1 0.0183 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Linear mixed model fit by maximum likelihood . t-tests use Satterthwaite's
## method [lmerModLmerTest]
## Formula: firstfixdur ~ distance + (1 | subject)
## Data: competitor
##
## AIC BIC logLik deviance df.resid
## -13761.5 -13733.3 6884.8 -13769.5 8577
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.0810 -0.4620 -0.1796 0.1467 8.1358
##
## Random effects:
## Groups Name Variance Std.Dev.
## subject (Intercept) 0.002044 0.04521
## Residual 0.011516 0.10731
## Number of obs: 8581, groups: subject, 55
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 1.108e-01 6.313e-03 5.899e+01 17.56 <2e-16 ***
## distanceshort 5.468e-03 2.317e-03 8.526e+03 2.36 0.0183 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## distancshrt -0.184##
## Descriptive statistics by group
## group: linear
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4292 0.12 0.12 0.08 0.09 0.05 0.02 1 0.98 3.91 18.77 0
## ------------------------------------------------------------
## group: nested
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4289 0.11 0.11 0.08 0.09 0.05 0.02 1 0.98 3.99 20.2 0##
## Descriptive statistics by group
## group: long
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4291 0.11 0.11 0.08 0.09 0.05 0.02 1 0.98 4.05 20.64 0
## ------------------------------------------------------------
## group: short
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4290 0.12 0.12 0.09 0.09 0.05 0.02 1 0.98 3.86 18.47 0Last run fixation duration on the competitor visual
In contrast to the general pattern, both the structure and distance improved the last run fixation model.
There were shorter fixations the on the competitor visual when processing nested sentences as compared to linear sentences. As in the general viewing pattern, there were longer fixations on the competitor visual when processing the shorter sentences as compared to longer sentences.
## Data: competitor
## Models:
## z0: lastfixdur ~ (1 | subject)
## z1: lastfixdur ~ structure + (1 | subject)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## z0 3 -10194 -10173 5100.2 -10200
## z1 4 -10198 -10169 5102.7 -10206 5.0465 1 0.02468 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Linear mixed model fit by maximum likelihood . t-tests use Satterthwaite's
## method [lmerModLmerTest]
## Formula: lastfixdur ~ structure + (1 | subject)
## Data: competitor
##
## AIC BIC logLik deviance df.resid
## -10197.5 -10169.2 5102.7 -10205.5 8577
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.3794 -0.6070 -0.1927 0.4443 6.4159
##
## Random effects:
## Groups Name Variance Std.Dev.
## subject (Intercept) 0.006485 0.08053
## Residual 0.017364 0.13177
## Number of obs: 8581, groups: subject, 55
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 1.924e-01 1.104e-02 5.688e+01 17.418 <2e-16 ***
## structurenested -6.392e-03 2.845e-03 8.526e+03 -2.247 0.0247 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## structrnstd -0.129## Data: competitor
## Models:
## z0: lastfixdur ~ (1 | subject)
## z2: lastfixdur ~ distance + (1 | subject)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## z0 3 -10194 -10173 5100.2 -10200
## z2 4 -10200 -10172 5103.9 -10208 7.4068 1 0.006498 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Linear mixed model fit by maximum likelihood . t-tests use Satterthwaite's
## method [lmerModLmerTest]
## Formula: lastfixdur ~ distance + (1 | subject)
## Data: competitor
##
## AIC BIC logLik deviance df.resid
## -10199.8 -10171.6 5103.9 -10207.8 8577
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.3262 -0.6110 -0.1902 0.4430 6.4150
##
## Random effects:
## Groups Name Variance Std.Dev.
## subject (Intercept) 0.006485 0.08053
## Residual 0.017360 0.13176
## Number of obs: 8581, groups: subject, 55
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 1.853e-01 1.104e-02 5.688e+01 16.778 <2e-16 ***
## distanceshort 7.745e-03 2.845e-03 8.526e+03 2.722 0.0065 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## distancshrt -0.129##
## Descriptive statistics by group
## group: linear
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4292 0.19 0.16 0.14 0.17 0.11 0.02 1 0.98 1.65 3.29 0
## ------------------------------------------------------------
## group: nested
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4289 0.19 0.15 0.14 0.16 0.11 0.02 1 0.98 1.69 3.56 0##
## Descriptive statistics by group
## group: long
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4291 0.19 0.15 0.14 0.16 0.11 0.02 1 0.98 1.78 4.09 0
## ------------------------------------------------------------
## group: short
## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 4290 0.19 0.16 0.14 0.17 0.12 0.02 1 0.98 1.57 2.86 0Working memory and competitor grid
There is no correlation between the working memory indicators and the percentage of time spent looking at the target visual.
##
## Pearson's product-moment correlation
##
## data: comp_wm_all$dtp and comp_wm_all$acc
## t = -0.80131, df = 53, p-value = 0.4265
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.3641352 0.1605500
## sample estimates:
## cor
## -0.1094072##
## Pearson's product-moment correlation
##
## data: comp_wm_all$dtp and comp_wm_all$rt
## t = 0.30071, df = 53, p-value = 0.7648
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.2265073 0.3032468
## sample estimates:
## cor
## 0.04126984##
## Pearson's product-moment correlation
##
## data: comp_wm_all$dtp and comp_wm_all$mspan
## t = -0.19071, df = 53, p-value = 0.8495
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.2894733 0.2407828
## sample estimates:
## cor
## -0.02618721##
## Pearson's product-moment correlation
##
## data: comp_wm_all$dtp and comp_wm_all$time
## t = 0.059706, df = 53, p-value = 0.9526
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.2576568 0.2729044
## sample estimates:
## cor
## 0.008200943Figures
Rainbow plots
ggsave(grid.arrange(p1,p2, ncol = 2),file="rainbowtarget.png", width = 9, height = 4)## Warning: Removed 3 rows containing missing values (`stat_slabinterval()`).## Warning: Removed 3 rows containing non-finite values (`stat_boxplot()`).## Warning: Removed 3 rows containing missing values (`stat_slabinterval()`).## Warning: Removed 3 rows containing non-finite values (`stat_boxplot()`).
ggsave(grid.arrange(p3,p4, ncol = 2),file="rainbowcomp.png", width = 9, height = 4)## Warning: Removed 3 rows containing missing values (`stat_slabinterval()`).## Warning: Removed 3 rows containing non-finite values (`stat_boxplot()`).## Warning: Removed 3 rows containing missing values (`stat_slabinterval()`).## Warning: Removed 3 rows containing non-finite values (`stat_boxplot()`).