An eye on semantics: a study on the influence of concreteness and predictability on early fixation durations

Figures & data

Figure 1. Intercorrelation matrix of relevant semantic, psycholinguistic, and context variables. All correlations greater than |.2| are reported numerically in the matrix.

Table 1. Examples of the stimuli, including the associated concreteness (range 1, highly abstract, to 5, highly concrete) and predictability (range 0, highly unpredictable, to 1, highly predictable).

Sentence	Concreteness (1–5)	Predictability (0–1)
After the knee injury, he needs to wear a brace when running.	3.96	0.64
There are plenty of clothes in the large basket donated to the charity shop.	5	0.25
Sarah is very meticulous and pays attention to every detail of her work.	2.5	0.96
Brokers need to detect any change in the mood of their clients.	1.75	0.14

Table 2. Summary statistics of psycholinguistics and semantic variables. In bold are highlighted the variables included in our analyses.

	Concreteness	Length	Bigram Frequency	Trigram Frequency	Orthographic Neighbour-hood Size	OLD20	Frequency	Zipf Value	Valence	Arousal	Sensorimotor strength
mean	3.25	5.51	21460.18	2344.83	3.34	1.86	5816.58	4.16	5.22	4.16	4.63
std	1.15	1.07	10782.10	2814.40	4.14	0.42	7390.54	0.55	1.35	1.00	0.93
min	1.19	4.00	3221.93	18.34	0.00	1.00	81.00	2.61	1.68	2.15	1.92
max	5.00	7.00	68055.67	28695.40	19.00	3.45	51028.00	5.40	8.05	7.05	7.22
	Age of Acquisition	Semantic Diversity	Context Similarity	PRECEDING Frequency	PRECEDING Zipf Value	PRECEDING Length	Position	Cloze Probability	Plausibility	Predictability
mean	7.86	1.70	0.15	2505782.39	5.92	4.64	8.46	0.28	5.78	0.43
std	2.32	0.25	0.08	3634761.17	1.42	2.73	1.55	0.32	0.70	0.28
min	2.5	0.74	0.01	11.00	1.77	1.00	6.00	0.00	2.80	0.02
max	15.56	2.29	0.55	9418422.00	7.67	13.00	12.00	1.00	6.77	1.00

Table 3. Summary of the measures, averaged across participants.

	FFD (ms)	GD (ms)	PrF (%)
Mean (SD)	220.97 (24.60)	235.81 (31.38)	68.08 (14.56)

Table 4. Results of the LME for the first fixation duration analysis.

Predictors	Estimates	CI	p
Intercept	220.7	212.98 – 228.41	<0.001
Frequency (Zipf)	-1.81	-3.93 – 0.30	0.093
Position	2.04	0.17 – 3.91	0.033
Length	0.32	-1.45 – 2.09	0.723
Predictability	-5.26	-7.18 − -3.33	<0.001
Concreteness	-2.07	-4.03 − -0.11	0.038
Predictability*Concreteness	-0.71	-2.47 – 1.05	0.429
Random Effects
σ^2	4198.86
τ00 ID	145.51
τ00 Subject.4	587.96
τ11 Subject.ConcM	6.8
τ11 Subject.Sim	5.17
τ11 Subject.Position	5.53
τ11 Subject.LogFreqZipf	15.18
ρ01
ρ01
ICC	0.03
NSubject	40
NID	400
Observations	10966
Marginal R^2 / Conditional R^2	0.011 / 0.046

Significant predictors (α=0.05) have been highlighted with p-value in bold.

Table 5. Results of the LME for the gaze duration analysis.

Predictors	Estimates	CI	p
Intercept	234.99	225.10–244.89	<0.001
Frequency (Zipf)	-3.1	-5.98 − -0.21	0.035
Length	4.94	1.90 – 7.99	0.001
Position	2.42	-0.13 – 4.97	0.062
Predictability	-8.64	-11.50 − -5.78	<0.001
Concreteness	-1.22	-4.06 – 1.63	0.402
Predictability*Concreteness	-0.64	-3.14 – 1.86	0.615
Random Effects
σ^2	7103.8
τ00 ID	346.53
τ00 Subject.5	955.29
τ11 Subject.ConcM	17.14
τ11 Subject.Sim	17.15
τ11 Subject.Position	4.89
τ11 Subject.LEN	29.93
τ11 Subject.LogFreqZipf	22.64
ρ01
ρ01
ICC	0.05
NSubject	40
NID	400
Observations	10966
Marginal R^2 / Conditional R^2	0.018 / 0.066

Significant predictors (α=0.05) have been highlighted with p-value in bold.

Table 6. Results for the probability of fixation logistic regression analysis.

Predictors	Odds Ratios	CI	p
Intercept	2.71	2.12 – 3.48	<0.001
Frequency (Zipf)	0.94	0.90 – 0.99	0.017
Length	1.42	1.34 – 1.50	<0.001
Position	0.96	0.92 – 1.00	0.069
Predictability	0.92	0.88 – 0.96	<0.001
Concreteness	0.98	0.93 – 1.03	0.473
Predictability*Concreteness	1	0.95 – 1.04	0.839
Random Effects
σ^2	3.29
τ00 ID	0.04
τ00 Subject.3	0.62
τ11 Subject.ConcM	0.01
τ11 Subject.LEN	0.01
τ11 Subject.LogFreqZipf	0
ρ01
ρ01
ICC	0.01
NSubject	40
NID	400
Observations	15676
Marginal R^2 / Conditional R^2	0.041 / 0.055

Significant predictors (α=0.05) have been highlighted with p-value in bold.

Figure 2. (A) Visual representation of the estimates in the model. (B) Comparison between conditional effects of concreteness and frequency in the full FFD model, when keeping the other predictors constant.

Table 7. LME model including all predictors that significantly influenced FFD.

Predictors	Estimates	CI	p
Intercept	220.65	212.93 – 228.37	<0.001
Frequency (Zipf)	-1.71	-3.82 – 0.40	0.113
Position	2.03	0.16 – 3.89	0.033
Lenght	0.37	-1.39 – 2.13	0.678
Predictability	-3.84	-6.04 − -1.65	0.001
Concreteness	-2.61	-4.60 − -0.61	0.010
Plausibility	-2.51	-4.90 − -0.11	0.040
Predictability*Concreteness	-0.47	-2.23 – 1.29	0.600
Random Effects
σ^2	4191.65
τ00 ID	142.39
τ00 Subject.5	588.71
τ11 Subject.plausibility	11.68
τ11 Subject.ConcM	6.46
τ11 Subject.Sim	1.23
τ11 Subject.Position	5.54
τ11 Subject.LogFreqZipf	15.11
ρ01
ρ01
ICC	0.04
N Subject	40
N ID	400
Observations	10966
Marginal R^2 / Conditional R^2	0.012 / 0.047

Significant predictors (α=0.05) have been highlighted with p-value in bold.