Similarity between referents constrains the processing of contrastive focus during reading

Abstract

Contrastive focus implies a contrast between two elements. However, it is unclear whether and how any interplay between such a contrast and similarity between potentially contrasting elements might affect focus processing. Accordingly, we report an eye movement experiment investigating this issue. The experiment used a background story to introduce eight characters whose social identities were manipulated to be similar or dissimilar. Participants first read this background story, then a series of two-sentence discourses while their eye movements were recorded. Each discourse referred to two characters from the passage who had either similar or dissimilar identities, with one (the target character) either focused using the Chinese particle zhiyou (meaning only) or unfocused. The results showed a typical focus facilitation effect, such that target character names were processed more quickly when focused than unfocused. We also observed a main effect of the similarity/dissimilarly of characters and, crucially, an interaction between this variable and focus. This interaction was due to slower processing of a post-target region when the target character was focused and the two characters had similar rather than dissimilar identities, but no such effect when the target character was unfocused. The findings suggest that establishing a contrast between referents is effortful during reading when these have similar rather than dissimilar social identities and so are more difficult to differentiate. The distinctiveness of referents in a discourse context may therefore constrain the establishment of contrastive focus during reading. We discuss these findings in relation to current theories of focus interpretation.

Keywords

Contrastive focus eye movements during reading Chinese

Sentences like “Only LinYue bought a car” not only convey that LinYue bought a car but also emphasise the subject “LinYue,” indicating that this is the focus of the sentence (Paterson et al., 2007; Umbach, 2004). This use of only additionally implies that other persons did not buy a car, so specifies a contrast between “LinYue” and some other individuals. More formally, particles such as only often serve to indicate a contrast between the referent of a focused constituent and its alternatives (Halliday, 1967; Rooth, 1992; Umbach, 2004). However, the particle does not directly specify the nature of this contrast and this raises the question of what factors constrain the establishment of contrastive focus.

Contrasts, nevertheless, appear to be established quickly and relatively effortlessly during reading. For instance, when processing a sentence like “He saw that there was a safe with a new lock and a safe with an old lock,” readers can construct a contrast between the two safes (Altmann & Steedman, 1988). Moreover, they are able to use this contrastive information to guide the processing of a subsequent syntactic ambiguity, so that in a following sentence like “The burglar blew open the safe with the new lock/with the dynamite . . .,” they process “with” as the beginning of a noun-phrase rather than a verb-phrase attachment (so modifying “the safe” rather than “blew open”). Other research additionally shows that the contrast implied by a focus particle like only is computed sufficiently rapidly that it can influence syntactic processing (Filik et al., 2005; Liversedge et al., 2002; Ni et al., 1996; Paterson et al., 1999; Sedivy, 2002). For instance, Ni et al. (1996) found that the inclusion of only could eliminate difficulty experienced when processing ambiguous sentences like, “Only businessmen loaned money at low interest were told to record their expenses,” Here only implies a contrast between two sets of businessmen, leading the reader to anticipate modifying information that specifies the nature of this contrast. As such information is supplied by the phrase “loaned money at low interest,” this is taken as a modifier of “the businessmen,” so that readers process this phrase as a relative clause, avoiding the garden path effect associated with this type of structure (Bever, 1970). However, for a sentence like “Only wealthy businessmen loaned money at low interest were told to record their expenses,” the requirement for modifying information can be satisfied by the adjective “wealthy” so that a garden path is still experienced by readers.

However, in some cases, contrasts fail to be established. Chen and Yang (2015) provided direct evidence that contrastive focus is more difficult to process when a contrast set is not easily available. In this study, a subject was focused using a cleft structure (i.e., “it was . . . who . . .,” using the particle shi in the original Chinese materials). Such sentences could not be processed easily when a contrast alternative for the focused word was absent, such as in “After Cong Yan arrived at the park, it was Cong Yan who bought the tickets” compared with an example where a contrast alternative was present, as in “After Zhen Zhou and Cong Yan arrived at the park, it was Cong Yan who bought the tickets.” Overall, the findings suggest that processing of contrastive focus is successful only when a contrast element is present in the sentence or discourse context (Chen, 2018; Chen & Yang, 2015) or easily inferred using world knowledge (Ni et al., 1996).

Further research suggests that processing is difficult when a supplied contrast has a different grammatical status to the focused constituent. For example, when reading sentences like, “At dinner, Jane passed only the salt to her mother but not the pepper/her father . . .,” processing difficulty is experienced at “her father” but not “the pepper.” This is because the focused phrase “the salt” has the same grammatical status as “the pepper” and so can more easily form a contrast, whereas it has a different grammatical status from “her father,” making contrast establishment more difficult (Paterson et al., 2007). A subsequent study by Fraundorf et al. (2013) provided additional evidence that the computation of contrast sets depends on their availability in the discourse model. Participants read short discourses that mentioned three same-category referents such as “Jupiter,” “Saturn,” and “Neptune,” in which “Jupiter” and “Saturn” were mentioned together in a sentence, while “Neptune” was mentioned in another sentence. A subsequent sentence mentioned “Jupiter” either normally or using capitals (e.g., “JUPITER”) to produce emphasis or not. In a subsequent probe recognition task, this use of capitals to produce emphasis facilitated the correct rejection of a sentence containing the referent “Saturn” but not “Neptune,” suggesting that only the more plausible discourse referent was encoded as a focus alternative.

Such findings raise the question of which constraints must be satisfied for readers to establish contrastive focus. Umbach’s (2004) analysis of contrast in information structures might help answer this question. Umbach proposed that contrastive elements should be comparable, presupposing the concepts of “similarity” and “dissimilarity.” Similarity means that two elements should have a “common integrator.” For instance, in a sentence like “John only bought the beer, not the martini,” the focus “beer” and the contrast “martini” belong to the superordinate category “drink,” which is the common integrator. The role of this constraint has been demonstrated in previous studies where readers had difficulty establishing a contrast between elements in different categories (Paterson et al., 2007). Moreover, this supposes that individuals would predict similar alternatives for contrast. For example, after listening to a context sentence like “Neil has some pears and some oranges,” listeners would look more at the picture of “apple” among four pictures once they heard a following sentence like “Alex only has some . . .,” compared with listening to a context sentence like “Neil has some sandals and some boots” (Kim et al., 2015). Accordingly, both Umbach’s theory and previous research findings suggest that some similarity between elements is required for contrastive focus and that integration difficulty is experienced when this is absent. However, Umbach’s theory also supposes that some dissimilarity between the two elements is required for focus establishment. In particular, the theory requires that the contrasting elements should be independent. If not, and one element subsumes the other (e.g., “drink” is the focus and “martini” is the alternative, or vice versa), a contrast cannot be established between them. Nevertheless, independence is only a minimum requirement to reach a plausible interpretation of sentences containing contrastive focus and empirical research is required to explore the boundary of dissimilarity between the focus and its alternatives. Therefore, the present study aimed to explore further the requirement that contrasting elements be dissimilar under conditions in which the minimum requirements that these have a common integrator but are also independent are satisfied.

To investigate the impact of dissimilarity on the processing of contrastive focus, we created social identities for several discourse referents (i.e., characters in a story). These could be discriminated in terms of age (young vs. middle-aged) and gender (male vs. female). Consequently, it was possible to create contrasts between characters that were either the same (i.e., similar) or different (i.e., dissimilar) on these dimensions. The logic of the approach is that, because notions of identity require consideration of both similarity and differences between people and things (Jenkins, 2014), this might be compatible with contrast processing. Importantly, specifying a contrast between two persons (with either similar or different social identities) fulfils the minimum requirement of the similarity constraint, as these belong to the same superordinate category (i.e., they are both humans). It also fulfils the minimum dissimilarity requirement as the two are different persons. This therefore allowed us to investigate whether additional degrees of similarity/dissimilarity between referents, as provided by age and gender information, would influence contrast establishment. Our intention with this manipulation was to gain further insight into the role of similarity/dissimilarity in the establishment of contrast sets.

Accordingly, we constructed a background story that introduced several characters in their social roles (Table 1) and a set of short discourses that referred to two of the characters and specified a contrast between them using the focus particle zhiyou or not (Table 2). Previous research has shown that the processing of a focused element is facilitated by the presence of a focus particle (i.e., the “focus facilitation effect,” Birch & Rayner, 2010; Chen et al., 2012; Morris & Folk, 1998). We therefore predicted that the presence of zhiyou would facilitate the processing of a focused target name as in previous eye movement research (Chen et al., 2019). This particle also elicits contrastive focus. For this to be successful, it seems likely that processes of differentiation and comparison will be activated (Xing & Yu, 2006). If similarity/dissimilarity in social identities between two characters can affect the establishment of contrastive focus online, we may observe an interaction effect in which reading times for a focused, but not unfocused, character shows sensitivity to the similarity/dissimilarity of the potential contrasting referents. In particular, if dissimilarity between referents facilitates focus establishment, reading times for a focused character’s name may be longer when the referents are similar rather than dissimilar in terms of their social roles, but with no such difference when the character’s name is unfocused. More exploratory analyses may also help reveal the time course of this effect and whether it emerges early during processing of the target names or affects only the later integration of this information. If it is delayed, the effect may appear later in the eye movement record or may be delayed so that they appear downstream of the target name in the sentence. However, if similarity/dissimilarity does not affect the online establishment of contrastive focus, this interaction may not be observed during either target name processing or later in the sentence.

Table 1.

Summary of background story.

TangGuo (male) and ZhangXiu (female) were preparing a birthday party for their daughter, XiaoHong. XiaoHong got back home from the university. She was ecstatic and invited her good friend XiaoMing (male) and XiaoMing’s parents, LinYue (male) and LiaoLi (female), to attend her birthday party. The two families had known each other for years. XiaoHong also invited XiaoHua (male), who was also a good friend of XiaoMing. Then XiaoHong invited her best friend, XiaoYing (female), and other friends to the birthday party. They prepared gifts for XiaoHong. That day, everyone went to XiaoHong’s house and celebrated XiaoHong’s birthday.

Table 2.

Examples of experimental passages.

Condition	Sentences
Similar-identity—focused	林岳和唐国一起去爬泰山，最后只有林岳登上了山顶。 LinYue and TangGuo climbed Taishan Mountain together. At last, only LinYue reached the summit.
Similar-identity—unfocused	林岳和唐国一起去爬泰山，最后林岳登上了山顶。 LinYue and TangGuo climbed Taishan Mountain together. At last, LinYue reached the summit.
Dissimilar-identity—focused	林岳和小红一起去爬泰山，最后只有林岳登上了山顶。 LinYue and XiaoHong climbed Taishan Mountain together. At last, only LinYue reached the summit.
Dissimilar-identity—unfocused	林岳和小红一起去爬泰山，最后林岳登上了山顶。 LinYue and XiaoHong climbed Taishan Mountain together. At last, LinYue reached the summit.

The target region is shown in bold font (e.g., “LinYue”) and the post-target region is underlined (e.g., “reached”), although text was shown normally in the experiment.

Method

Participants

We recruited 48 female undergraduates aged 18–25 years (M_age = 20 years) with normal or corrected vision from Zhejiang Normal University. Participants were paid 20–30 Yuan for taking part. Participants were intentionally selected to be female to reduce the potential for sex differences in the processing of social information (Newman et al., 2008) and focus (Wang et al., 2011). The study was approved by the Animal Ethical and Welfare Committee of FJNU and conducted in accordance with the principles of the Declaration of Helsinki. All participants gave informed written consent.

Stimuli and design

The experimental materials were a background story that was read before the experiment, and 48 sets of two-sentence mini discourses read during the experiment. The background story was about celebrating one character’s birthday and contained 1,580 Chinese characters (see Table 1 for a summary of the story and Supplemental Appendix 1 for the whole story). The story described a series of events that occurred for eight fictional characters, including two middle-aged male professionals (TangGuo and LinYue), two middle-aged female professionals (LiaoLi and ZhangXiu), two young male undergraduate students (XiaoMing and XiaoHua), and two young female undergraduate students (XiaoHong and XiaoYing). This permitted a manipulation of similarity/dissimilarity by gender and age.

Forty-eight sets of two-sentence mini discourses were constructed, each with four versions (see Table 2). The first sentence of each mini discourse introduced two characters with either a similar social identity (e.g., “LinYue and TangGuo”) or dissimilar social identities (e.g., “LinYue and XiaoHong”) and described an event involving these characters (e.g., “climbed Taishan Mountain together”). The second sentence described a subsequent event (e.g., “reached the summit”), undertaken by one of the characters, whose name (e.g., “LinYue” in the second sentence) was either focused by placing the particle zhiyou before it or not. For half of the mini discourses, the first-mentioned name in the first sentence was repeated as the target name, and for the other mini discourses the second-mentioned name was repeated as the target. The experiment therefore had a within-participants design that manipulated the factors Focus (focused, unfocused) and Similarity/Dissimilarity (similar, dissimilar).

The mini discourses were divided into four lists so that only one version of each mini discourse appeared in each list, with each list containing an equal number of mini discourses in each condition. Each list contained 106 stimuli, comprising one version of each of the 48 experimental mini discourses (12 per condition), an additional 48 filler passages, and 10 practice passages. The filler and practice passages were also two-sentence mini discourses, with the number of the characters, form of co-reference, and similarity between the characters varied. The particle zhiyou was used in only six filler items, always emphasising information other than a repeated name.

Apparatus and procedure

Participants were assessed individually. Each was first asked to read the background story. They were then given five comprehension questions to confirm they understood the story and could remember the identities of the eight main characters. If a participant made any errors in answering these questions, they were asked to reread the story. Once the participant could answer all questions correctly, they proceeded to the eye-tracking experiment.

At the start of the eye movement experiment, the participant was informed of the task and instructed to read for comprehension. The participant was then seated in a height-adjustable chair with head movements minimised using a chin rest. An EyeLink 1000 eye-tracker (SR Research Inc., Canada) was used to record each participant’s right eye movements during binocular reading. This was calibrated using a 9-point procedure to ensure the average spatial accuracy within 0.5°. Stimuli were displayed as white text on a black background in 24-point Song font using the UMass EyeTrack software (version 0.7.10m; http://blogs.umass.edu/eyelab/software/). At the viewing distance in the experiment, each character subtended approximately 1°.

Each trial began with a drift correction followed by a small white square that appeared on the left side of the screen. Once the participant fixated this square, it disappeared and was replaced by the first character of the passage. Once the participant finished reading a passage, they pressed a button on a response pad. The passage then disappeared, replaced by a yes/no comprehension question on one third of trials. Participants responded to this by pressing one of two buttons on the response pad. The experiment lasted about 30 min for each participant.

Results

One participant’s data was excluded from analyses because of poor accuracy responding to the comprehension questions that followed passage presentations (i.e., performance less than 80% correct). Average accuracy answering the questions was 97.4% for the remaining 47 participants (range = 82%–100%). Prior to data analysis, fixations shorter than 80 ms were merged into larger fixations within one character, after which fixations shorter than 80 ms and longer than 500 ms were deleted (using UMass EyeDoctor software, version 0.6.5). This affected 3.6% of fixations.

Data are reported for a target region comprising the target name in the second sentence of each passage (e.g., “LinYue”) and a post-target region comprising the two characters following the target name to capture any spillover effects (e.g., “bought”). We computed six standard eye movement measures for these regions (see Rayner, 2009). These included measures sensitive to first-pass processing of text, which is the initial processing within an interest region prior to a saccade to the right of this region or a regression from this region. First-pass measures used were first-fixation duration (duration of the first fixation within a region), and gaze duration (sum of all first-pass fixations within a region). As measures of later processing, we examined regressions-in (the percentage probability of a regression back to a region) and second-pass reading time (sum of all fixations from the first regression to the region prior to a saccade from the region). Finally, as measures of overall processing, we examined the total number of fixations within a regions and total reading time (sum of all fixations within a region). Trials in which the region was skipped were excluded from first-fixation duration and gaze duration analyses. For each measure, ANOVAs by participants and by items were conducted, treating participants (F₁) and items (F₂) as random factors. The significant level was set at alpha = .05.

Table 3 shows mean eye movement measures and Tables 4 and 5 show summarised statistical analyses for the target and post-target regions.

Table 3.

Mean eye movement measures for target and post-target regions.

	Target region		Post-target region
	Similar	Dissimilar	Similar	Dissimilar
	M (SE)	M (SE)	M (SE)	M (SE)
First-fixation duration (ms)
Focused	247 (5.5)	244 (5.5)	249 (7.0)	233 (5.5)
Unfocused	245 (5.8)	250 (5.5)	241 (6.3)	241 (4.8)
Gaze duration (ms)
Focused	257 (5.9)	257 (7.0)	268 (8.7)	249 (7.3)
Unfocused	270 (6.5)	275 (6.9)	259 (7.9)	259 (7.0)
Total reading time (ms)
Focused	312 (18.1)	290 (15.5)	301 (18.4)	268 (14.2)
Unfocused	344 (19.9)	336 (18.6)	320 (18.4)	331 (19.2)
Second-pass time (ms)
Focused	148 (15.8)	122 (12.2)	98 (11.7)	79 (9.5)
Unfocused	174 (15.5)	165 (16.1)	123 (14.4)	118 (13.9)
Total number of fixations
Focused	1.39 (.08)	1.29 (.06)	1.37 (.08)	1.23 (.06)
Unfocused	1.53 (.09)	1.48 (.08)	1.47 (.09)	1.41 (.08)
Regressions-in (%)
Focused	41 (2.9)	35 (2.6)	26 (2.3)	20 (2.3)
Unfocused	51 (3.1)	52 (3.4)	27 (2.4)	25 (2.7)

M: mean; SE: standard error.

Table 4.

Statistical results for the target region.

Measures	Factors	F₁ analysis				F₂ analysis
Measures	Factors	MSE	F	p	η²	MSE	F	p	η²
First-fixation duration (ms)	S/D	25	0.03	.867	.001	488	0.71	.404	.015
	F	241	0.38	.538	.008	408	0.45	.506	.009
	S/D × F	847	1.49	.228	.031	2,394	2.88	.096	.058
Gaze duration (ms)	S/D	405	0.27	.609	.006	41	0.03	.873	.001
	F	11,525	7.98	.007**	.148	13,450	6.68	.013*	.124
	S/D × F	301	0.32	.573	.007	2,738	1.88	.177	.038
Total reading time (ms)	S/D	10,847	2.17	.148	.045	13,200	2.17	.147	.044
	F	71,643	12.80	<.001***	.218	65,490	13.48	<.001***	.223
	S/D × F	2,621	0.77	.384	.017	4,602	0.66	.420	.014
Second-pass time (ms)	S/D	13,515	3.93	.053	.079	13,686	4.04	.050	.079
	F	55,425	11.56	.001**	.201	56,478	12.06	.001**	.204
	S/D × F	3,320	1.02	.318	.022	4,209	0.95	.334	.020
Total number of fixations	S/D	0.27	3.31	.075	.067	0.25	2.50	.121	.050
	F	1.28	12.35	.001**	.212	1.10	14.05	<.001***	.230
	S/D × F	0.03	0.36	.550	.008	0.07	0.64	.429	.013
Regressions-in (%)	S/D	405	1.91	.174	.040	453	2.01	.162	.041
	F	8,338	24.68	<.001***	.349	7,069	25.15	<.001***	.349
	S/D × F	629	2.86	.097	.059	504	1.45	.235	.030

MSE: mean squared error; F: focus; S/D: similarity/dissimilarity.

The degrees of freedom for the participants analysis (F₁) was 1, 46 and for the items analysis (F₂) was 1, 47.

p < .05, **p < .01, ***p < .001.

Table 5.

Statistical results for post-target region.

Measures	Factors	Subject analysis				Item analysis
Measures	Factors	MSE	F	p	η²	MSE	F	p	η²
First-fixation duration (ms)	S/D	3,137	3.45	.070	.070	838	0.73	.396	.015
	F	0.02	.000	.996	.000	417	0.40	.531	.008
	S/D × F	3,040	6.48	.014*	.123	3,897	3.16	.082	.063
Gaze duration (ms)	S/D	4,232	2.18	.146	.045	1,776	0.81	.373	.017
	F	19	0.01	.922	.000	300	0.19	.661	.004
	S/D × F	4,581	4.69	.035*	.093	4,351	2.22	.143	.045
Total reading time (ms)	S/D	5,578	1.33	.255	.028	5,601	0.85	.361	.018
	F	78,108	21.16	<.001***	.315	72,191	16.80	<.001***	.263
	S/D × F	21,747	5.35	.025*	.104	19,421	5.15	.028*	.099
Second-pass time (ms)	S/D	6,986	3.13	.084	.064	7,450	2.08	.156	.042
	F	46,793	18.06	<.001***	.282	46,252	15.34	<.001***	.246
	S/D × F	2,402	1.76	.192	.037	3,367	1.15	.289	.024
Total number of fixations	S/D	0.46	5.83	.020*	.113	0.36	3.09	.085	.062
	F	0.94	14.74	<.001***	.243	0.71	11.19	.002**	.192
	S/D × F	0.07	0.98	.327	.021	0.03	0.39	.533	.008
Regressions-in (%)	S/D	881	4.79	.034*	.094	721	5.27	.026*	.101
	F	514	3.40	.071	.069	403	1.70	.199	.035
	S/D × F	167	1.31	.258	.028	243	1.89	.176	.039

MSE: mean squared error; F: focus; S/D: similarity/dissimilarity.

The degree of freedom for the participants analysis (F₁) was (1, 46), and for the items analysis (F₂) it was (1, 47).

p < .05, **p < .01, ***p < .001.

Target region

At the target region, there was a main effect of Focus on gaze durations, regressions-in, second-pass reading times, total number of fixations, and total reading times. This was due to shorter reading times, fewer fixations, and fewer regressions for focused than non-focused names. As these effects were observed in both first-pass (i.e., gaze durations) and later measures of processing, it appears that focus affected both the initial and later processing of words, consistent with evidence from previous research (Birch & Rayner, 2010; Chen et al., 2012; Morris & Folk, 1998). No other effects were observed at this region.

Post-target region

At the post-target region, there was a main effect of Focus on second-pass reading times, total number of fixations, and total reading times. This was due to shorter reading times and fewer fixations in this region following focused than non-focused names. These effects are consistent with a “spillover” of the focus effect from the target region.

There was also a main effect of Similarity/Dissimilarity in regressions-in and total number of fixations (significant by participants only), reflecting more fixations and regressions back to the post-target region when the characters had similar rather than dissimilar social identities. Crucially, there was an interaction between Focus and Similarity/Dissimilarity (see Figure 1) in first-fixation durations (significant by participants), gaze durations (significant by participants), and total reading times (significant by both participants and items). Pair-wise comparisons showed no reading time differences within the post-target region following non-focused target names with similar or dissimilar social identities (Fs < 1). However, reading times were longer following focused words with similar rather than dissimilar social identities (first-fixation durations, F₁(1, 46) = 6.60, p = .014, F₂(1, 47) = 3.89, p = .054; gaze durations, F₁(1, 46) = 5.17, p = .028, F₂(1, 47) = 3.14, p = .083; total reading times, F₁(1, 46) = 5.54, p = .023, F₂(1, 47) = 4.98, p = .030). These findings suggest that processing of contrastive focus is sensitive to the similarity/dissimilarity of potential contrasting elements. Moreover, it appears that while this influenced focus processing, effects emerge relatively late in the eye movement record.

Figure 1.

Mean total reading times at the post-target region. The target name was either focused or unfocused and the two characters in the text had either similar to dissimilar social identities. Error bars represent the standard error of the mean.

Recently, von der Malsburg and Angele (2017) proposed that false positives can be elevated when computing multiple dependent measures in eye movement studies. Accordingly, we adopted two approaches to control for an inflated risk of a Type I error due to assessing multiple eye movement measures across two regions of interest. First, by applying a Bonferroni correction that took account of the multiple dependent measures across the two regions, we reset the critical alpha level to .004 (i.e., an alpha of .05 divided by 12 dependent measures). After applying this correction, we still observed significant focus effects in second-pass reading times, total reading times, and regressions-in at the target word region (with a marginal effect in gaze durations), confirming findings from previous research. However, effects of similarity/dissimilarity and the interaction between focus and similarity/dissimilarity at the post-target region fell short of this adjusted significance level. A second approach to reducing false positives is to consider if an effect is statistically significant (at α = .05) in two or more measures. Following this alternative criterion, the main effect of focus (in gaze durations, second-pass reading times, total reading times and regressions-in) and similarity/dissimilarity (in total number of fixation and regressions-in), and the interaction between focus and similarity/dissimilarity (in first-fixation duration, gaze duration, and total reading time), would be reliable. We therefore note that, based on these analyses, both methods produce statistically reliable focus effects and the latter approach provides support for an interaction between focus and the similarity/dissimilarity of the contrasted referents.

Discussion

The present experiment investigated the influence of focus and the similarity/dissimilarity of potential contrasting elements on eye movements during reading. The findings showed a clear focus facilitation effect, due to focused words receiving shorter reading times than non-focused words. This effect was robust even under stringent correction of alpha levels to avoid false positives (following von der Malsburg & Angele, 2017), and in line with previous research showing that focusing devices like “only” or “zhiyou” can cause attentional resources to be allocated to a focused element, enabling this to be processed more efficiently (Birch & Rayner, 2010; Chen et al., 2012; Chen & Yang, 2015; Morris & Folk, 1998). As this focus effect was observed in measures sensitive to both the initial and later processing of words, it appears that focus facilitated the identification of a character’s name as well as the subsequent processing of this referent. Moreover, as this effect was observed in a non-alphabetic script (Chinese) in the present experiment, whereas previous effects have been reported primarily for alphabetic scripts such as English, it appears that this aspect of focus processing is similar across alphabetic and non-alphabetic writing systems.

Crucially, the experiment also produced an interaction between focus and referent similarity/dissimilarity in reading times (weakly in first-fixation and gaze durations, and robustly in total reading times) at a post-target region containing the following word. No effect of similarity/dissimilarity was observed when the target name was unfocused. However, when the particle zhiyou was present, reading times were longer in similar than dissimilar identity conditions. The findings, therefore, provide support for the constraint of dissimilarity on the processing of contrastive focus. While this effect was not reliable under stringent correction of alpha levels, it was statistically significant (at α = .05) in multiple measures, suggesting it is not a false positive resulting from multiple comparisons (von der Malsburg & Angele, 2017). In formulating hypotheses, we were unable to predict whether this effect would emerge early or late in the eye movement record. Therefore, the finding that it emerges late, in a post-target region, and so likely reflects sentence integration processes, will help future research test more specific hypotheses concerning the nature and timing of such effects.

According to the theory of information structure (Halliday, 1967; Umbach, 2004), focus cues the establishment of a contrast between a focused referent and its alternatives. In the present case, it seems that the establishment of contrast between a focused character and its alternative was more difficult when these had similar rather than dissimilar social roles. We believe that this finding can be interpreted in terms of Umbach’s (2004) hypothesis that contrast is due to similarity plus dissimilarity. That is, the contrastive elements should share some similar features and possess some different features, so that they can be contrasted. As a minimum requirement, this can be operationalised in terms of the two elements having a common integrator. Previous research shows that when this constraint is violated (such that the elements do not have a common integrator) this can lead to processing difficulty (Drenhaus et al., 2011; Paterson et al., 2007). With the present experiment, we examined a situation in which contrasted elements had a common integrator (they were both human) and so satisfied this minimum requirement. This enabled us to test the further requirement from Umbach’s theory that elements that meet this criterion must also have some minimum difference between them to facilitate contrast establishment. In the present case, they had either similar or dissimilar identities, specified in terms of age or gender. We found that when potential contrasts had similar identities (and so lacked a dissimilarity), readers found it more difficult to establish contrastive focus. Accordingly, the present findings build on previous focus processing research by showing that this depends on both components of Umbach’s theory and that readers have difficulty establishing a contrast between elements that meet minimum requirements for similarity but lack a dissimilarity. As the incurred processing cost appeared late in the eye movement record, affecting processing in a post-target region, it appears to reflect computational processes that occur relatively late during focus establishment. The present findings nevertheless demonstrate that both dissimilarity and similarity are important for establishing contrastive focus.

In addition to these effects, we observed a main effect of dissimilarity in regression and total number of fixations for the post-target region, suggesting that readers looked back and made more regressions to this region in similar versus dissimilar social identity conditions independent of the manipulation of focus. This suggests that even without a focus particle, reference to a focused name was more difficult in the context of a potential contrast with a similar rather than dissimilar social identity. This was unpredicted. However, a possible explanation for this effect is that the two discourse characters were inclined to be contrasted even in the absence of an explicit focus marker. In previous research, Altmann and Steedman (1988) showed that when two safes were introduced as “a safe with a new lock and a safe with an old lock,” an implicit contrast between the two was established, even though readers were not instructed to do so. This suggested that implicit contrasts can be specified in the absence of overt focus marking. Consequently, while we observed pervasive effects of contrast establishment difficulty in eye movements when sentences included a focus particle, weaker effects may also have been observed in the absence of this particle, due to the processing of an implicit contrast. Accordingly, while it will be important to investigate further which factors constrain the processing of contrastive focus, it also may be valuable to conduct this research in relation to the processing of both explicit and implicit contrasts.

In sum, the present provides new insight into the establishment of contrastive focus online during reading by showing that this is influenced by the similarity/dissimilarity of potential contrast sets. Such findings provide empirical support for Umbach’s (2004) theory that “contrast due to similarity plus dissimilarity” guides the processing of contrast sets, adding to our understanding of mechanisms underlying the establishment of contrastive focus when reading for comprehension.

Supplemental Material

QJE-STD-20-111.R1-Supplementary_Material – Supplemental material for Similarity between referents constrains the processing of contrastive focus during reading

Supplemental material, QJE-STD-20-111.R1-Supplementary_Material for Similarity between referents constrains the processing of contrastive focus during reading by Shuang Chen, Yuqing Tang, Xiuna Lv, Kevin B Paterson and Lijing Chen in Quarterly Journal of Experimental Psychology

Footnotes

Acknowledgements

The authors thank Yuejuan Wang for her help in collecting the data.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by a grant from the National Natural Science Foundation of China to L.C. (Grant No. 31500916), a 1000 Talents Visiting Professorship to Kevin Paterson, and a grant from the National Natural Science Foundation of China to S.C. (Grant No. 31700990).

ORCID iDs

Kevin B Paterson

Lijing Chen

Supplementary Material

The Supplementary Material is available at

References

Altmann

G. T. M.

Steedman

(1988). Interaction with context during human sentence processing. Cognition, 30(3), 191–238.

Bever

T. G.

(1970). The cognitive basis for linguistic structures. Cognition and the Development of Language, 279(362), 1–61.

Birch

Rayner

(2010). Effects of syntactic prominence on eye movements during reading. Memory & Cognition, 38, 740–752.

Chen

(2018). On the processing of pragmatic information: ERP effects of emphasis processing. Language, Cognition and Neuroscience, 33(8), 1005–1016.

Chen

Yang

(2012). Focus, newness and their combination: Processing of information structure in discourse. PLOS ONE, 7(8), e42533.

Chen

Paterson

Yang

(2019). Pragmatic influences on sentence integration: Evidence from eye movements. Quarterly Journal of Experimental Psychology, 72(12), 2742–2751.

Chen

Yang

(2015). Emphasizing the only character: Emphasis, attention and contrast. Cognition, 136, 222–227.

Drenhaus

Zimmermann

Vasishth

(2011). Exhaustiveness effects in clefts are not truth-functional. Journal of Neurolinguistics, 24(3), 320–337.

Filik

Paterson

K. B.

Liversedge

S. P.

(2005). Parsing with focus particles in context: Eye movements during the processing of relative clause ambiguities. Journal of Memory and Language, 53(4), 473–495.

10.

Fraundorf

S. H.

Benjamin

A. S.

Watson

D. G.

(2013). What happened (and what didn’t): Discourse constraints on encoding of plausible alternatives. Journal of Memory and Language, 69(3), 196–227.

11.

Halliday

M. A. K.

(1967). Notes on transitivity and theme in English: Part 2. Journal of Linguistics, 3(2), 199–244.

12.

Jenkins

(2014). Social identity. Routledge.

13.

Kim

C. S.

Gunlogson

Tanenhaus

M. K.

Runner

J. T.

(2015). Context-driven expectations about focus alternatives. Cognition, 139, 28–49.

14.

Liversedge

S. P.

Paterson

K. B.

Clayes

E. L.

(2002). The influence of only on syntactic processing of “long” relative clause sentences. Quarterly Journal of Experimental Psychology Section A: Human Experimental Psychology, 55(1), 225–240.

15.

Morris

R. K.

Folk

J. R.

(1998). Focus as a contextual priming mechanism in reading. Memory & Cognition, 26(6), 1313–1322.

16.

Newman

M. L.

Groom

C. J.

Handelman

L. D.

Pennebaker

J. W.

(2008). Gender differences in language use: An analysis of 14,000 text samples. Discourse Processes, 45(3), 211–236.

17.

Crain

Shankweiler

(1996). Sidestepping garden paths: Assessing the contributions of syntax, semantics and plausibility in resolving ambiguities. Language and Cognitive Processes, 11, 283–334.

18.

Paterson

K. B.

Liversedge

S. P.

Filik

Juhasz

B. J.

White

S. J.

Rayner

(2007). Focus identification during sentence comprehension: Evidence from eye movements. Quarterly Journal of Experimental Psychology, 60, 1423–1445.

19.

Paterson

K. B.

Liversedge

S. P.

Underwood

(1999). The influence of focus operators on syntactic processing of short relative clause sentences. Quarterly Journal of Experimental Psychology Section A: Human Experimental Psychology, 52(3), 717–737.

20.

Rayner

(2009). Eye movements and attention in reading, scene perception, and visual search. Quarterly Journal of Experimental Psychology, 62(8), 1457–1506.

21.

Rooth

(1992). A theory of focus interpretation. Natural Language Semantics, 1(1), 75–116.

22.

Sedivy

J. C.

(2002). Invoking discourse-based contrast sets and resolving syntactic ambiguities. Journal of Memory and Language, 46(2), 341–370.

23.

Umbach

(2004). On the notion of contrast in information structure and discourse structure. Journal of Semantics, 21(2), 155–175.

24.

Von der Malsburg

Angele

(2017). False positives and other statistical errors in standard analyses of eye movements in reading. Journal of Memory and Language, 94, 119–133.

25.

Wang

Bastiaansen

Yang

Hagoort

(2011). The influence of information structure on the depth of semantic processing: How focus and pitch accent determine the size of the N400 effect. Neuropsychologia, 49(5), 813–820.

26.

Xing

(2006). Social comparison: Contrast effect or assimilation effect? Advances in Psychological Science, 14(6), 944–949. (In Chinese)

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.04 MB