Clauses that are parallel in form and meaning show processing advantages in ellipsis and coordination structures (
The online interpretation of ellipsis structures has become a popular topic among psycholinguists because it highlights an intriguing mismatch between form and meaning, and consequently reveals a unique demand on the human sentence processing system. In particular, that a meaningful interpretation is recovered from ellipsis shows that it is not enough for the processor to simply parse linguistic structure by passively interpreting the word forms presented to it; instead, the processor must actively go beyond the input and infer the correct form at the appropriate level of representation, e.g., syntactic, Logical Form, discourse, etc. We aim to expand the empirical and conceptual coverage of ellipsis processing by exploring an understudied ellipsis type known as
Although there are many types of ellipsis structures, perhaps the most well known case is VP (verb phrase) ellipsis, which we use to illustrate the basic inference problem faced by the human language sentence processor. For example, in (1a) the auxiliary
(1) | a. | John ate a cheeseburger. Bill did, too. |
b. | John ate a cheeseburger. Bill ate a cheeseburger, too. |
To interpret (1a) as (1b), the processor must therefore “fill in” the missing or elided material, presumably by consulting linguistic or discourse representations from the context. Current research on ellipsis suggests that the processor engages in some kind of cost-free mechanism, retrieving the missing form through either copying (
(2) | John ate a cheeseburger. Bill did <eat a cheeseburger>, too. |
If the problem of recovering the ellipsis weren’t thorny enough, the processor may also need to infer that additional material or content is missing. In cases of clausal ellipsis like sluicing (
(3) | a. | John ate something, but I don’t know [CP what1 <John ate |
b. | John ate, but I don’t know [CP what1 <John ate |
While there are various syntactic accounts of sluicing (e.g.,
Processing sprouting has been shown to elicit online processing costs for sluicing ellipsis (
(4) | a. | The secretary typed (something), but I don’t know what. | |
b. | The secretary typed (somewhere), but I don’t know where. |
However, processing a conjunction structure is independently facilitated when the conjuncts are parallel in their syntactic, semantic, or prosodic structure (e.g.,
(5) | a. | |
b. |
The existing evidence thus suggests that structural parallelism
Focus-sensitive coordination structures are a class of constructions that contain a coordinator like
(6) | a. | I can’t drink beer, much less vodka. |
b. | I can’t drink beer, much less [FocP vodka]1 <I drink |
Though not reviewed in detail here, the ellipsis account of focus-sensitive coordination structures is supported by many distributional tests that align it with so-called small conjunct ellipsis (see
Focus-sensitive coordination is similar to stripping/bare argument ellipsis and sluicing ellipsis in that the correlate and the remnant bear pitch accent in the most felicitous pronunciation, typically with accents that mark contrastive focus (
(7) | a. | I can’t drink BEER, much less VODKA. |
b. | I can’t DRINK beer, much less MAKE it. | |
c. | I can’t DRINK BEER, much less MAKE WINE. | |
d. | I can’t drink ONE beer, much less TWO (beers). |
However, research on a range of ellipsis structures has found that overt prosodic marking on a noun increases the likelihood of it being selected as a correlate, but does not entirely disambiguate the sentence (e.g.,
Finally, and perhaps most crucially for present purposes, the antecedent and the elided clause stand in a scalar relationship. In particular, the negative antecedent (
The second kind of scales are sometimes known as
It is currently unclear whether one kind of scale might be more difficult to recover than the other. Conventionalized scales might be psychologically privileged, in that they could be accessible in more general contexts than
Example (8) illustrates that the (contextually) weaker element in focus-sensitive coordination must be located in the first conjunct;
(8) | a. | John didn’t steal SOME of the cookies, much less ALL of them. |
b. | #John didn’t steal ALL of the cookies, much less SOME of them. | |
c. | John didn’t steal ALL of the cookies, but/although he did steal SOME of them. |
The basic properties we assume for focus-sensitive coordination are summarized in (9).
(9) | ||
1. | The second conjunct consists of the remnant of ellipsis; | |
2. | The correlate and remnant are usually marked with contrastive focus; and | |
3. | Propositions formed from the clauses containing the correlate and remnant are placed on a contextually salient scale. |
We believe that these properties place a unique set of demands on the sentence processing system. As we shall see, they also allow us to formulate two potentially opposing hypotheses regarding the processing routines recruited to interpret ellipsis structures in real time. First, however, we articulate our assumptions regarding what basic information the processor needs in order to process focus-sensitive coordination generally.
As the literature on processing ellipsis structures is rapidly growing (e.g.,
(10) | ||
1. | Parse the remnant of the ellipsis, i.e., construct the appropriate phrase structure for the remnant given the input. | |
2. | Locate the correlate, if any, from the antecedent clause. | |
3. | Construct or infer the elided phrase, e.g., by regenerating or copying a structure at Logical Form. |
We use example (6) to illustrate the three tasks in more depth in (11). We assume that each process depends on the previous one. For instance, the parser must have established the basic syntactic category of the remnant (step 1) before it can locate a correlate of the appropriate type in the antecedent clause (step 2). Similarly, generating a structure for the elided clause (step 3) depends on having selected an appropriate correlate from the antecedent clause (step 2).
(11) | I can’t drink beer, let alone vodka. |
Step 1. |
|
I can’t drink beer, much less [DP=Remnant vodka]. | |
Step 2. |
|
suitable contrast to the remnant |
|
I can’t drink [DP=Correlate beer ], much less [DP=Remnant vodka]. | |
Step 3. |
|
I can’t drink [DP=Correlate beer ], much less [DP=Remnant vodka] 1 < I drink |
Although focus-sensitive coordination has been studied far less than other kinds of ellipsis structures, there are a few recent results that bear on the processes outlined above. Regarding step 1, Harris (
With respect to step 2, Harris & Carlson (
What, then, is the processor to do when the clause containing the remnant and ellipsis is not parallel with the antecedent? Instances of sprouting discussed above represent an extreme case, in which the remnant lacks a constituent in the antecedent. As reviewed above, prior studies have found that sprouted correlates incur a processing cost during online comprehension, due either to an ellipsis-specific operation in which the ellipsis site is modified to include a variable corresponding to the correlate (
In example (12a), the processor presumably must pair the remnant
(12) | ||
a. | Michael couldn’t study carpentry, much less chemistry. | |
b. | Michael couldn’t study, much less chemistry. |
In sluicing ellipsis, however, the relationship between clauses is much different. In the case with sprouting (13b), no such entailment relationship holds between the antecedent clause (
(13) | ||
a. | Michael studied something, but I don’t know what. | |
b. | Michael studied, but I don’t know what. |
In the remainder of this paper, we explore two possibilities. On the one hand, it is in principle conceivable that the cost for sprouting is limited to sluicing and other types of ellipsis besides focus-sensitive coordination, in that, without a scalar relation between clauses, there would be no interpretive advantage for sprouting in these structures. That is, if
(14) |
This possibility rests on the premise that
On the other hand, it is possible that the processor always prefers a correlate that is maximally parallel to the remnant as a matter of course, regardless of the advantage to interpretation. We define parallelism as the presence of any of a number of similarities (morphological, prosodic, syntactic, semantic, etc.) between contrasting or conjoined phrases (
(15) |
As noted earlier, initial studies of parallelism mainly concentrated on unelided conjoined elements instead of ellipsis. Frazier et al. (
In various studies of ambiguous ellipsis sentences, Carlson (
We take parallelism to be an extra-grammatical factor that affects processing of ellipsis structure and conjoined structures, rather than a grammatical constraint. The most extreme nonparallel condition is one in which a remnant lacks a correlate entirely, as in sprouting examples of sluicing or focus-sensitive coordination structures, which are nonetheless entirely grammatical structures. We take these cases as a testing ground for comparing the predictions of the Scalar Advantage Principle and the Parallel Contrast Principle.
When it comes to sprouting in focus-sensitive coordination, the two principles make the opposite predictions. The Scalar Advantage Principle predicts an advantage for sprouting, and the Parallel Contrast Principle predicts a penalty for sprouting. There is already some evidence for the Parallel Contrast Principle for focus-sensitive coordination. Carlson & Harris (
(16) | a. | I don’t own [a hat], much less [a red one]. |
b. | She will not argue with [a fool], much less [a money-hungry one]. (COCA) |
A series of auditory and written questionnaires confirmed that zero-adjective contrast was dispreferred compared to examples with parallel DPs in naturalness rating tasks, and was avoided in sentence completion. Finally, in a self-paced reading study with items like (17), they observed a penalty immediately on DP remnants (
(17) | a. | The chef didn’t overcook |
b. | The chef didn’t |
Despite the existing evidence that zero-adjective contrasts are costly to compute, the following set of experiments follows up on a number of remaining questions and concerns. First, we are cautious about analogizing the addition of an adjective to the remnant too closely with the established case of sprouting in sluicing. Where the analogy breaks down is that the remnant DP does in fact have a correlate DP in the antecedent clause in such cases, but the correlate simply lacks the expected sub-contrast within it. Thus, the processor might not need to posit a variable at logical form so much as readjust the kinds of scales or comparisons it can accommodate. Second, Carlson & Harris’ (
We present three norming studies and two eye tracking studies below, each of which contain items based on the following pattern (18). The first three conditions have PP (prepositional phrase) remnants, whereas the last three conditions have VP remnants (18d–f), which both served as a statistical control and prohibited participants from anticipating a PP remnant. The first sentence (18a) illustrates PP sprouting, in which a PP remnant completely lacks a corresponding PP correlate (the No Matrix PP condition). The second sentence (18b) contains an overt PP correlate that matches the remnant along syntactic and pragmatic dimensions (both indicate time; the Compatible PP condition). The third (18c) provides a case of moderate correlate-remnant mismatch (the Incompatible PP condition). While the correlate is of the same syntactic type, a PP, it doesn’t permit comparison along a comparable scale (the remnant is about time while the supposed correlate is about location). Bold formatting was added here and elsewhere for clarity of exposition, but did not appear in the experiment.
(18) | ||
a. | John doesn’t want to eat out, much less |
|
b. | John doesn’t want to eat out |
|
c. | John doesn’t want to eat out |
|
d. | John doesn’t want to eat out, much less |
|
e. | John doesn’t want to eat out |
|
f. | John doesn’t want to eat out |
The second eye tracking study explores whether the effects of PP sprouting are mitigated in supporting contexts, comparing the effect of sprouting cases like (18a) over controls (18b) with and without contexts introducing the PP remnant. In all, the results support the Parallel Contrast Principle, in that readers rely heavily on the form of the antecedent clause to identify suitable correlates, and that this process persists even in contexts supporting the PP remnant.
Materials were created by truncating the sentences to be used in the first eye tracking study (22) after
(19) | a. | John doesn’t want to eat out, much less _____________. |
b. | John doesn’t want to eat out |
|
c. | John doesn’t want to eat out |
The 30 experimental items were interspersed with 48 fragments from unrelated experiments, 15 non-experimental filler items, and 5 highly constrained fill-in-the blank fillers (e.g.,
Forty native speakers of English were recruited on Amazon’s Mechanical Turk, and were compensated $2 for participation. Four participants were removed for not passing three difficult-to-parse questions that demanded a native or highly proficient understanding of English. All remaining participants provided appropriate responses to the highly constrained catch items. Seven participants were removed for counterbalancing purposes, resulting in 27 total participants distributed equally across 3 counterbalancing lists. Data from six items that did not appear in the eye tracking study below were removed from the analysis. Results from the entire data set of 30 items set are virtually identical to the results presented here.
Both authors annotated the responses according to which element in the matrix clause the completion contrasted with, regardless of the syntactic category of the completion. For example, both
Two effects are clearly observable from Table
Experiment 1A: Completion norming study. Percentage of completions supplied by subjects by grammatical category.
Completion contrast | |||
---|---|---|---|
Matrix | PP | DP | VP |
No Matrix PP | 4% | 40% | 56% |
Compatible PP | 78% | 8% | 14% |
Incompatible PP | 81% | 9% | 10% |
Experiments 1A–B. Norming studies. Left panel: Results from the sentence completion study. Right panel: Results from the naturalness rating study presented as centered z-scores.
Second, in cases where there was no PP in the matrix clause, participants completed sentence fragments largely with either VP (56%) or DP (40%) completions,
To determine how similar completions to experimental items were to patterns observed in text and speech, we annotated 1670 relevant examples of the
(20) | There are few national markets to simplify the distribution of water |
(21) | There ought not to be violent crimes, much less |
Overall, the results from the corpus and completion studies show that PP sprouting is avoided in production. We now examine whether sprouting a PP remnant is penalized when provided directly to a comprehender.
Thirty sextets of sentence items were created, crossing Matrix clause structure (
Forty-nine participants were recruited on Amazon’s Mechanical Turk, and paid $2 for participation. Participants were identified as unique from those in the previous norming experiment by two criteria: their Amazon Worker ID number, and an anonymous variant of their IP address. One individual self-identified as a non-native speaker of English and was removed from the dataset. Thirteen others were removed for rating any of the ungrammatical catch items as a 4 or above, and four more were removed for counterbalancing purposes. The final dataset consisted of data from 30 participants equally distributed across 6 lists.
The data are presented in Table
Experiment 1B: Naturalness ratings. Uncorrected/z-score normalized means. Standard errors are in parentheses.
Matrix | PP penalty | ||
---|---|---|---|
No Matrix PP | 5.00/–0.42 (0.14) | 5.63/0.03 (0.13) | 0.63 |
Compatible PP | 5.89/0.22 (0.11) | 5.71/0.09 (0.12) | –0.18 |
Incompatible PP | 5.46/–0.09 (0.13) | 5.84/0.18 (0.11) | 0.38 |
Experiment 1B: Naturalness ratings. Linear mixed effects regression model. Parameters with
Parameter | Estimate | Std. Error | |
---|---|---|---|
(Intercept) | 5.589 | 0.156 | 35.86* |
No Matrix PP | –0.26 | 0.059 | –4.41* |
Incompatible PP | 0.044 | 0.059 | 0.74 |
PP Remnant | –0.145 | 0.042 | –3.47* |
Incompatible PP × PP Remnant | –0.056 | 0.059 | –0.95 |
No Matrix PP × PP Remnant | –0.157 | 0.059 | –2.66* |
Items from the No Matrix PP condition (
According to the Scalar Advantage Principle, PP sprouting should be preferred as a way to avoid computing
Items consisted of 24 items (22) from the naturalness experiment, half of which were followed by comprehension questions; see Appendix A for a complete list. Items were interspersed with another 48 sentences from two unrelated experiments and 18 non-experimental filler items. Prior to analysis, sentences were partitioned into 6 regions. As the No Matrix PP conditions (22a) did not contain a PP region or any other linguistic content in that portion of the matrix clause, the PP region was coded as empty (🚫) for analysis.
(22) | a. | /John doesn’t want / to eat out/ 🚫 /, much less/{ |
b. | /John doesn’t want / to eat out on Saturday,/ much less /{ |
|
c. | /John doesn’t want / to eat out at a steakhouse,/ much less /{ |
The number of characters in PP remnants (
Participants were instructed to read silently and at their own pace, and were given a short practice session to illustrate the procedure. The reader’s head was stabilized with a tower mount of an SR Research Eyelink 1000 eye tracker, which sampled eye movements from the right eye at 1000 Hz. Viewing was binocular. The display monitor was situated 55 cm away from the subject. All items were presented on a single line in 13-point fixed-width (proportional) Monaco font on a 21” LCD monitor using a Lenovo computer to display the sentences, so that three characters subtended approximately 1 degree of visual angle. Participants were calibrated before the experiment began with a three-point calibration system, and eye movement drift was corrected manually between each trial. Participants were encouraged to take breaks as often as they wished, and were calibrated if they moved away from the tower or if their fixations became unstable. A game pad was used to record responses to comprehension questions like (23) appearing after approximately half to the sentences. Comprehension questions contained either yes-no responses or simple forced-choice options. Access to the Internet was turned off on all computers, as were all non-essential programs.
(23) | Does John want to stay in instead of going out? | |
a. | Yes | |
b. | No |
Sixty UCLA undergraduates were recruited through the Psychology Pool for course credit. If a participant blinked on the remnant region in first pass reading on more than 3 trials for any one condition, she was removed from the data, and another participant was run in her place under the same counterbalancing list. Linguistic history was recorded for all subjects, all of whom self-reported as native speakers of English. All participants had normal or corrected to normal vision.
We report significant results for several standard eye tracking measures:
All statistical models were given the same deviation coding and random effects structures as the naturalness ratings experiment. Following standard practice, models of continuous measures used the Gaussian distribution, whereas models of binomial data were modeled as logistic linear regressions. Prior to analysis, outliers from first fixation and first pass distributions were censored using winsorsation, so that the scores below the 5th percentile and above the 95th percentile are replaced with the score at the 5th and 95th percentile, respectively (
Only the remnant and the region following are reported for first fixation, first pass, go past times, and regressions out; see Table
Experiment 1C: Eye tracking Means and standard errors for first fixation durations, first pass times, go past times, regressions out on the Remnant and Spill over regions.
Contrast | Remnant | Region | Region | ||
---|---|---|---|---|---|
No Matrix PP | PP Remnant | 216 (4) | 226 (4) | 424 (12) | 226 (4) |
VP Remnant | 216 (4) | 229 (4) | 476 (17) | 229 (4) | |
Compatible PP | PP Remnant | 222 (4) | 228 (3) | 426 (13) | 228 (3) |
VP Remnant | 219 (4) | 225 (4) | 480 (16) | 225 (4) | |
Incompatible PP | PP Remnant | 223 (4) | 227 (4) | 438 (13) | 227 (4) |
VP Remnant | 232 (5) | 226 (4) | 502 (15) | 226 (4) | |
No Matrix PP | PP Remnant | 584 (23) | 620 (26) | 20% (3) | 4% (1) |
VP Remnant | 592 (25) | 572 (25) | 14% (2) | 4% (1) | |
Compatible PP | PP Remnant | 488 (18) | 558 (24) | 9% (2) | 1% (1) |
VP Remnant | 607 (25) | 567 (24) | 17% (3) | 3% (1) | |
Incompatible PP | PP Remnant | 558 (24) | 593 (25) | 17% (3) | 4% (1) |
VP Remnant | 623 (25) | 565 (24) | 14% (2) | 3% (1) |
Experiment 1C: Eye tracking. Means and standard errors for regressions in, second pass, and total times for all regions.
Contrast | Remnant | ||||||
---|---|---|---|---|---|---|---|
No Matrix PP | PP Remnant | 17% (3) | — | 23% (3) | 6% (2) | 30% (3) | — |
VP Remnant | 17% (3) | — | 15% (3) | 4% (1) | 31% (3) | — | |
Compatible PP | PP Remnant | 26% (3) | 8% (2) | 11% (2) | 4% (1) | 29% (3) | — |
VP Remnant | 24% (3) | 6% (2) | 18% (3) | 5% (2) | 31% (3) | — | |
Incompatible PP | PP Remnant | 23% (3) | 8% (2) | 18% (3) | 8% (2) | 32% (3) | — |
VP Remnant | 26% (3) | 12% (2) | 17% (3) | 5% (2) | 24% (3) | — | |
No Matrix PP | PP Remnant | 113 (20) | — | 88 (11) | 60 (12) | 133 (18) | — |
VP Remnant | 87 (14) | — | 63 (10) | 41 (9) | 113 (15) | — | |
Compatible PP | PP Remnant | 167 (28) | 45 (9) | 44 (9) | 39 (9) | 120 (16) | — |
VP Remnant | 150 (22) | 46 (8) | 55 (8) | 45 (10) | 113 (14) | — | |
Incompatible PP | PP Remnant | 162 (27) | 57 (11) | 60 (9) | 49 (10) | 130 (16) | — |
VP Remnant | 133 (21) | 67 (11) | 55 (8) | 58 (11) | 102 (16) | — | |
No Matrix PP | PP Remnant | 1305 (37) | — | 402 (16) | 568 (22) | 710 (28) | 609 (25) |
VP Remnant | 1243 (38) | — | 381 (14) | 586 (22) | 649 (26) | 649 (25) | |
Compatible PP | PP Remnant | 1320 (38) | 529 (21) | 364 (13) | 501 (21) | 654 (26) | 625 (25) |
VP Remnant | 1296 (40) | 549 (23) | 380 (13) | 595 (25) | 658 (26) | 620 (24) | |
Incompatible PP | PP Remnant | 1351 (41) | 513 (22) | 364 (12) | 549 (20) | 680 (26) | 620 (25) |
VP Remnant | 1314 (39) | 523 (21) | 365 (13) | 629 (25) | 637 (25) | 585 (21) |
Experiment 1C: Eye tracking. Linear mixed effects regression models for the remnant and spill over regions for first fixation durations, first pass times, go past times, and percentage of regressions out.
Region | Parameters | ||||||
---|---|---|---|---|---|---|---|
Estimate | Std. Error | Estimate | Std. Error | ||||
(Intercept) | 248.64 | 11.15 | 22.3* | 111.38 | 29.13 | 3.82* | |
PP Remnant | –1.62 | 2.26 | –0.72 | –23.26 | 4.60 | –5.06* | |
Incompatible PP | 7.15 | 3.19 | 2.24* | 12.96 | 6.47 | 2.00* | |
No Matrix PP | –6.23 | 3.19 | –1.95+ | –6.60 | 6.48 | –1.02 | |
Length | –1.25 | 0.60 | –2.10* | 20.44 | 1.51 | 13.54* | |
PP Remnant × Incompatible PP | –4.52 | 3.19 | –1.42 | –5.49 | 6.48 | –0.85 | |
PP Remnant × No Matrix PP | 1.72 | 3.20 | 0.54 | 2.85 | 6.48 | 0.44 | |
(Intercept) | 231.96 | 5.09 | 45.57* | 231.96 | 5.09 | 45.57* | |
PP Remnant | –0.19 | 1.98 | –0.09 | –0.19 | 1.98 | –0.09 | |
Incompatible PP | 0.20 | 2.80 | 0.07 | 0.20 | 2.80 | 0.07 | |
No Matrix PP | 0.16 | 2.80 | 0.06 | 0.16 | 2.80 | 0.06 | |
PP Remnant × Incompatible PP | 1.25 | 2.81 | 0.45 | 1.25 | 2.81 | 0.45 | |
PP Remnant × No Matrix PP | –3.04 | 2.81 | –1.08 | –3.04 | 2.81 | –1.08 | |
(Intercept) | 178.52 | 48.52 | 3.68* | –1.02 | 0.43 | –2.35* | |
PP Remnant | –24.27 | 8.22 | –2.95* | 0.06 | 0.12 | 0.50 | |
Incompatible PP | 16.06 | 11.57 | 1.39 | 0.16 | 0.12 | 1.38 | |
No Matrix PP | 12.17 | 11.59 | 1.05 | –0.01 | 0.09 | –0.07 | |
Length | 23.44 | 2.58 | 9.09* | –0.06 | 0.02 | –2.58* | |
PP Remnant × Incompatible PP | –0.88 | 11.58 | –0.08 | 0.13 | 0.12 | 1.11 | |
PP Remnant × No Matrix PP | 27.65 | 11.60 | 2.38* | 0.25 | 0.12 | 2.16* | |
(Intercept) | 572.57 | 40.87 | 14.01* | –3.61 | 0.29 | –12.33* | |
PP Remnant | 6.29 | 7.81 | 0.81 | 0.10 | 0.24 | 0.43 | |
Incompatible PP | 0.38 | 11.03 | 0.03 | 0.29 | 0.23 | 1.25 | |
No Matrix PP | 12.53 | 11.03 | 1.14 | –0.03 | 0.17 | –0.17 | |
PP Remnant × Incompatible PP | 2.07 | 11.07 | 0.19 | 0.33 | 0.24 | 1.36 | |
PP Remnant × No Matrix PP | 9.69 | 11.05 | 0.88 | 0.01 | 0.23 | 0.03 |
Experiment 1C: Eye tracking. Linear mixed effect regression models for regressions in and total times.
Region | Parameters | ||||||||
---|---|---|---|---|---|---|---|---|---|
Estimate | Std. Error | Wald Z | Estimate | Std. Error | |||||
Subject | (Intercept) | –1.54 | 0.18 | –8.69* | 1308.8 | 62.71 | 20.87* | ||
PP Remnant | –0.01 | 0.07 | –0.14 | 17.44 | 11.89 | 1.47 | |||
Incompatible PP | 0.15 | 0.10 | 1.49 | 26.15 | 16.82 | 1.55 | |||
No Matrix PP | –0.34 | 0.11 | –3.12* | –27.45 | 16.79 | –1.64 | |||
PP Remnant × Incompatible PP | –0.09 | 0.10 | –0.82 | 3.35 | 16.84 | 0.20 | |||
PP Remnant × No Matrix PP | 0.01 | 0.11 | 0.11 | 3.19 | 16.80 | 0.19 | |||
PP Region | (Intercept) | –2.48 | 0.62 | –4.00* | 98.32 | 42.97 | 2.29* | ||
PP Remnant | –0.10 | 0.14 | –0.72 | –5.80 | 8.58 | –0.68 | |||
Incompatible PP | –0.19 | 0.14 | –1.37 | 7.47 | 8.69 | 0.86 | |||
Length | –0.02 | 0.04 | –0.59 | 28.21 | 2.50 | 11.28* | |||
PP Remnant × Incompatible PP | 0.18 | 0.14 | 1.34 | 4.04 | 8.57 | 0.47 | |||
Much less | (Intercept) | –1.89 | 0.17 | –11.4* | 374.67 | 13.54 | 27.68* | ||
PP Remnant | 0.06 | 0.11 | 0.52 | 1.20 | 4.86 | 0.25 | |||
Incompatible PP | 0.14 | 0.11 | 1.21 | –11.02 | 6.88 | –1.60 | |||
No Matrix PP | 0.02 | 0.08 | 0.25 | 14.47 | 6.87 | 2.11* | |||
PP Remnant × Incompatible PP | 0.02 | 0.11 | 0.15 | 1.43 | 6.88 | 0.21 | |||
PP Remnant × No Matrix PP | 0.28 | 0.11 | 2.50* | 9.95 | 6.88 | 1.45 | |||
Remnant | (Intercept) | –3.09 | 0.21 | –14.82* | 116.00 | 46.98 | 2.47* | ||
PP Remnant | 0.10 | 0.13 | 0.77 | –24.73 | 7.41 | –3.34* | |||
Incompatible PP | 0.22 | 0.18 | 1.24 | 17.21 | 10.43 | 1.65 | |||
No Matrix PP | –0.12 | 0.20 | –0.60 | 6.43 | 10.43 | 0.62 | |||
Length | — | — | — | 26.98 | 2.42 | 11.14* | |||
PP Remnant × Incompatible PP | 0.13 | 0.18 | 0.75 | –8.75 | 10.44 | –0.84 | |||
PP Remnant × No Matrix PP | 0.15 | 0.19 | 0.79 | 23.38 | 10.44 | 2.24* | |||
Spill over | (Intercept) | –1.10 | 0.23 | –4.85* | 664.8 | 47.28 | 14.06* | ||
PP Remnant | 0.06 | 0.07 | 0.79 | 13.37 | 7.91 | 1.69 | |||
Incompatible PP | –0.09 | 0.10 | –0.88 | –5.93 | 11.19 | –0.53 | |||
No Matrix PP | 0.06 | 0.10 | 0.59 | 12.03 | 11.18 | 1.08 | |||
PP Remnant × Incompatible PP | 0.19 | 0.10 | 1.92+ | 2.20 | 11.21 | 0.20 | |||
PP Remnant × No Matrix PP | –0.06 | 0.10 | –0.62 | 12.33 | 11.19 | 1.10 | |||
Final | (Intercept) | — | — | — | 612.5 | 35.75 | 17.13* | ||
PP Remnant | — | — | — | 0.63 | 7.47 | 0.08 | |||
Incompatible PP | — | — | — | –12.27 | 10.55 | –1.16 | |||
No Matrix PP | — | — | — | 12.50 | 10.55 | 1.18 | |||
PP Remnant × Incompatible PP | — | — | — | 12.54 | 10.57 | 1.19 | |||
PP Remnant × No Matrix PP | — | — | — | –17.11 | 10.56 | –1.62 |
Compatible with the results from the norming studies, an advantage for PP remnants over VP remnants appeared on the remnant region for first pass (a 57 ms advantage), go past (a 64 ms advantage), and total times (also 64 ms advantage); see the top left panel of Figure
Experiment 1C. Top left panel: PP Remnant advantage on the remnant region for first pass, go past, and total time measures. Remaining panels: Interaction between Remnant type and Matrix clause structure showing that the advantage for PP remnants is reduced or overturned in No Matrix PP conditions.
Crucially, the predicted exception to the advantage for PP remnants occurred when there was no PP in the matrix clause for the remnant to contrast with, i.e., just in the case of PP Sprouting, shown in Figure
A model with the first and second halves of the experiment included as an interactive predictor provided a better fit of the go past data on the remnant than other models computed. In this model, the interaction between Remnant type and the Incompatible PP condition was significantly reduced in the second half (13 ms penalty for Incompatible PP remnants over VP remnants) compared to the first half of the experiment (a 73 ms penalty for Incompatible conditions),
The results indicate three effects of primary interest. First, the processor appeared to encounter less processing difficulty on PP remnants than on VP remnants, suggesting an overall preference to form a contrast with the immediately preceding PP rather than the matrix VP that contained it. This interpretation is supported by the results of the completion norming study, in which PP contrasts were in general supplied at a much greater rate than any other contrast. Second, the PP advantage failed to hold in the case of PP Sprouting, where PP remnants elicited a processing cost in multiple measures (go past, total times, and regressions out) immediately on the remnant region itself. As predicted by the Parallel Contrast Principle, parallelism between the matrix and ellipsis clause appeared to facilitate processing, even though the conditions with PPs in the first clause required the processor to form an
Third, PP remnants that formed a semantically incompatible relation with their PP correlate were penalized in go past times on the remnant, but only in the first half of the experiment. This pattern suggests that participants were initially sensitive to the meaning incongruence of PPs relating to different aspects of the situation. Later in the study, they appeared to adopt a reading strategy of ignoring this minor mismatch, most likely due to repeated exposure to examples as the experiment wore on. This is an important conceptual control, as it confirms that subjects were attuned to the implied relationships between correlate-remnant pairs, rather than simply finding a correlate for the remnant supplied without consideration of its meaning. If readers were taking the Incompatible PPs to be part of the antecedent clause that would need to be copied, but not contrasting directly with the remnant PPs, then this condition should have patterned with the No Matrix PP condition: both would involve no contrasting correlate at all for the remnant PP. The slight dispreference instead suggests that participants did consider the PPs in different clauses to be potential correlates, albeit not as parallel as they might have liked, and that they adjusted over the course of the experiment to the mismatch in semantic content.
However, as all of the sentences were presented without context, it’s possible that the penalty attributed to PP sprouting is due instead to the introduction of a new referential DP contained within the PP remnant. This possibility is addressed in the following eye tracking experiment by adding preceding contexts that either mentioned the PP, thus making the remnant more accessible in the discourse, or did not. If violating parallelism is the central reason behind the penalty for PP sprouting in the studies above, then sprouting should continue to be costly, regardless of context.
This norming experiment had the same task as Experiment 1A, except that the target sentence fragments were preceded by contexts that either supported a PP remnant (24a) or were neutral (24b). As before, we manipulated whether there was a PP in the matrix clause (
(24) | a. | |
b. | ||
Forty-two participants completed a completion norming study for course credit over the Internet. Five participants identified as non-native speakers of English and were removed from analysis, as were six participants who answered highly predictable catch items incorrectly. Three more participants were removed for counterbalancing purposes, leaving 28 participants in the final dataset, who contributed 560 completions in total.
Twenty-two ambiguous, unclear, or non-sensical completions were removed. In the remaining completions, sprouting of any category appeared in 12% of cases, of which 60% were PP remnants. When the target contained a PP correlate in the matrix clause, there was only one instance of sprouting of any kind. However, when the target did not contain a PP correlate, there were more PP sprouting completions, regardless of context (0% vs. 15%). There was also an interaction, so that supporting contexts produced nearly twice the number of PP sprouting completions when there was no PP correlate (11% vs. 19%) compared to neutral contexts. The completion study indicates that contexts meant to induce or facilitate PP sprouting indeed resulted in more PP sprouting, even though sprouting was still avoided in general.
Twenty quartets like (25) were created in a design that crossed Context (Supporting, Neutral) and PP contrast (Matrix PP, No Matrix PP). In all items, the target sentences derived from sentences in the first eye tracking experiment. A complete list of contexts and items appears in Appendix B. There were no incompatible PPs in this study, so all correlate PPs were on the same semantic scale as the remnants, and there were no VP remnants. The context conditions were minimally different from each other, varying mostly in whether the remnant in the target sentence was overtly mentioned: it appeared in the Supporting contexts but not the Neutral ones.
(25) | a. | |
b. | ||
Items were interspersed with 66 items from unrelated experiments, and 20 non-experimental fillers, for a total of 106 items per experimental session. Comprehension questions like (26) appeared after approximately half of the materials.
(26) | Who was surprised at Oliver’s interest in travel? | |
a. | His friends | |
b. | His family |
Fifty-six subjects participated in the experiment, using the same recruitment and exclusion criteria described in the previous eye tracking experiment (Experiment 1C).
The data cleaning and analysis procedure from Experiment 1C were used in the present experiment, except that conditions were sum coded so that Supporting context and the Matrix PP conditions were treated as the statistical reference levels. Means and standard errors are reported in Tables
Experiment 2B: Eye tracking in context. Means and standard errors for first fixation durations, first pass times, go past times, regressions out.
Context | Matrix PP | Remnant | Spill over | Remnant | Spill over |
---|---|---|---|---|---|
Supporting | Matrix PP | 216 (5) | 228 (5) | 340 (14) | 465 (21) |
No Matrix PP | 217 (4) | 228 (5) | 352 (14) | 433 (18) | |
Neutral | Matrix PP | 222 (5) | 229 (5) | 346 (13) | 438 (17) |
No Matrix PP | 218 (5) | 231 (5) | 381 (15) | 475 (19) | |
Supporting | Matrix PP | 415 (21) | 531 (31) | 12% (3) | 6% (2) |
No Matrix PP | 433 (22) | 480 (26) | 16% (3) | 5% (2) | |
Neutral | Matrix PP | 433 (20) | 506 (31) | 15% (3) | 3% (1) |
No Matrix PP | 487 (25) | 563 (33) | 18% (3) | 8% (2) |
Experiment 2B: Eye tracking in context. Means and standard errors for regressions in, second pass times, total times.
Context | Matrix PP | ||||||
---|---|---|---|---|---|---|---|
Supporting | Matrix PP | 13% (3) | 11% (3) | 15% (3) | 1% (1) | 34% (4) | — |
Supporting | No Matrix PP | 11% (3) | — | 19% (4) | 6% (2) | 34% (4) | — |
Neutral | Matrix PP | 18% (3) | 8% (2) | 17% (3) | 2% (1) | 32% (4) | — |
Neutral | No Matrix PP | 13% (3) | — | 20% (4) | 5% (2) | 29% (4) | — |
Supporting | Matrix PP | 110 (31) | 50 (12) | 41 (11) | 17 (6) | 124 (19) | — |
Supporting | No Matrix PP | 96 (27) | — | 62 (15) | 53 (14) | 126 (17) | — |
Neutral | Matrix PP | 118 (22) | 41 (10) | 52 (10) | 35 (10) | 107 (17) | — |
Neutral | No Matrix PP | 123 (27) | — | 77 (15) | 49 (13) | 110 (20) | — |
Supporting | Matrix PP | 1149 (47) | 456 (21) | 368 (17) | 385 (19) | 607 (32) | 498 (26) |
Supporting | No Matrix PP | 1095 (45) | — | 351 (20) | 439 (22) | 567 (28) | 484 (31) |
Neutral | Matrix PP | 1102 (51) | 459 (21) | 353 (15) | 431 (21) | 544 (25) | 488 (24) |
Neutral | No Matrix PP | 1106 (48) | — | 371 (22) | 496 (31) | 634 (36) | 541 (34) |
Experiment 2B: Eye tracking in context. Linear mixed effects models for first fixation durations, first pass times, go past, and regressions out.
Parameters | ||||||||
---|---|---|---|---|---|---|---|---|
Estimate | Std. Error | Estimate | Std. Error | |||||
(Intercept) | 218.02 | 3.52 | 62.01* | 349.77 | 19.19 | 18.23* | ||
Neutral | –2.41 | 2.25 | –1.07 | –10.53 | 5.53 | –1.91+ | ||
No Matrix PP | –0.35 | 2.24 | –0.16 | 11.61 | 5.52 | 2.10* | ||
Neutral × No Matrix PP | 1.78 | 2.27 | 0.78 | –1.28 | 5.60 | –0.23 | ||
(Intercept) | 228.98 | 4.14 | 55.33* | 433.06 | 26.88 | 16.11* | ||
Neutral | –1.56 | 2.22 | –0.70 | –2.90 | 7.10 | –0.41 | ||
No Matrix PP | 1.17 | 2.23 | 0.53 | 5.36 | 7.15 | 0.75 | ||
Neutral × No Matrix PP | 0.18 | 2.25 | 0.08+ | –11.81 | 7.19 | –1.64 | ||
(Intercept) | 438.32 | 21.46 | 20.43* | –2.02 | 0.22 | –9.04* | ||
Neutral | 18.22 | 10.43 | 1.75+ | 0.10 | 0.14 | 0.75 | ||
No Matrix PP | 15.85 | 10.41 | 1.52 | 0.12 | 0.14 | 0.90 | ||
Neutral × No Matrix PP | 7.35 | 10.50 | 0.70 | –0.02 | 0.14 | –0.13 | ||
(Intercept) | 505.45 | 33.58 | 15.05* | –3.81 | 0.55 | –6.89* | ||
Neutral | 12.13 | 13.55 | 0.90 | –0.10 | 0.24 | –0.42 | ||
No Matrix PP | 5.49 | 13.63 | 0.40 | 0.27 | 0.24 | 1.11 | ||
Neutral × No Matrix PP | 21.54 | 13.68 | 1.57 | 0.35 | 0.24 | 1.46 |
Experiment 2B: Eye tracking in context. Linear mixed effects regression models for regressions in and total times.
Region | Parameter | ||||||
---|---|---|---|---|---|---|---|
Estimate | Std. Error | Wald Z | Estimate | Std. Error | |||
(Intercept) | –2.08 | 0.21 | –9.83* | 1116.33 | 69.34 | 16.10* | |
Neutral | 0.15 | 0.14 | 1.11 | 4.02 | 18.94 | 0.21 | |
No Matrix PP | –0.18 | 0.14 | –1.30 | 10.01 | 18.88 | 0.53 | |
Neutral × No Matrix PP | –0.07 | 0.14 | –0.48 | 9.02 | 19.09 | 0.47 | |
(Intercept) | –2.24 | 0.22 | –10.42* | 450.21 | 23.86 | 18.87* | |
Neutral | –0.14 | 0.22 | –0.63 | 3.12 | 14.05 | 0.22 | |
(Intercept) | –1.83 | 0.23 | –7.85* | 357.01 | 13.51 | 26.42 | |
Neutral | 0.06 | 0.13 | 0.44 | 3.30 | 8.96 | 0.37 | |
No Matrix PP | 0.15 | 0.13 | 1.11 | 3.47 | 8.93 | 0.39 | |
Neutral × No Matrix PP | –0.03 | 0.13 | –0.20 | 7.00 | 8.99 | 0.78 | |
(Intercept) | –3.61 | 0.40 | –9.10* | 433.17 | 25.34 | 17.09* | |
Neutral | 0.20 | 0.32 | 0.61 | 26.41 | 11.06 | 2.39* | |
No Matrix PP | 0.74 | 0.32 | 2.31* | 29.41 | 11.04 | 2.66* | |
Neutral × No Matrix PP | –0.3 | 0.32 | –0.92 | –1.00 | 11.13 | –0.09 | |
(Intercept) | –0.81 | 0.15 | –5.41* | 571.77 | 39.06 | 14.64* | |
Neutral | –0.09 | 0.10 | –0.86 | –2.91 | 12.74 | –0.23 | |
PP Sprouting | –0.05 | 0.10 | –0.49 | 15.12 | 12.74 | 1.19 | |
Neutral × No Matrix PP | –0.04 | 0.10 | –0.34 | 22.91 | 12.83 | 1.79+ | |
(Intercept) | — | — | — | 492.66 | 36.16 | 13.62* | |
Neutral | — | — | — | 14.97 | 12.09 | 1.24 | |
No Matrix PP | — | — | — | 13.01 | 12.10 | 1.08 | |
Neutral × No Matrix PP | — | — | — | 12.05 | 12.2 | 0.99 |
Experiment 2B: Eye tracking in context. By-subjects and by-items ANOVAs for second pass times.
Region | Effect | By-subjects | By-items | ||
---|---|---|---|---|---|
Subject | Context | 0.73 | 0.40 | 0.44 | 0.52 |
Matrix PP | 0.22 | 0.64 | 0.00 | 0.98 | |
Context × Matrix PP | 0.00 | 0.96 | 0.02 | 0.88 | |
Matrix PP | Context | 0.17 | 0.68 | 0.18 | 0.67 |
Much less | Context | 0.05 | 0.83 | 1.16 | 0.30 |
Matrix PP | 2.21 | 0.14 | 2.75 | 0.11 | |
Context × Matrix PP | 1.35 | 0.25 | 0.06 | 0.80 | |
Remnant | Context | 0.03 | 0.86 | 0.28 | 0.61 |
Matrix PP | 4.49 | <0.05 | 4.35 | <0.05 | |
Context × Matrix PP | 2.54 | 0.12 | 0.93 | 0.35 | |
Spill over | Context | 1.16 | 0.29 | 0.37 | 0.55 |
Matrix PP | 0.15 | 0.70 | 0.22 | 0.65 | |
Context × Matrix PP | 0.09 | 0.77 | 0.14 | 0.71 |
There was a marginal 18 ms penalty for Neutral contexts in first pass times on the remnant,
Experiment 2B. Left panel: Elongated reading times for Neutral compared to Supporting contexts in first pass and total time measures in the remnant region. Right panel: Reading time penalty for PP Sprouting conditions in first pass, second pass, and total time measures on the remnant.
As in the previous experiment, sprouting in the No Matrix PP conditions was found to be costly immediately on the remnant region, regardless of preceding context, as shown in the right panel of Figure
While not central to the main hypotheses, we expected that Supporting contexts would facilitate recovery from any difficulty due to sprouting a PP argument. No interactions were observed in first fixation or first pass times. However, first pass times in which regressions out were eliminated indeed showed an interaction
Experiment 2B. Regression contingent analyses of first pass times on the remnant shown as centered z-scores. Left panel: First pass times in trials with regressions out of the remnant region showed a differential effect of context on No Matrix PP conditions, but no effect on Matrix PP conditions. Right panel: First pass times in trials without regressions out of the remnant region showed an advantage for Supporting contexts.
Other measures showed weak evidence in favor of a reduced PP sprouting penalty in Supporting contexts. Although not the best fitting model, an interaction consistent with a PP sprouting penalty was observed in go past times on the post remnant spill over region once trial order was included as an interactive predictor; there was a 57 ms cost for the No Matrix PP condition in the Neutral contexts, but a 51 ms advantage for PP Sprouting in Supporting contexts
The results confirm the findings from the previous eye tracking experiment: there was a reading time penalty for PP remnants that lacked an overt correlate, as predicted by the Parallel Contrast Principle. Although the cost for PP sprouting appeared in a variety of measures regardless of the context, it was reduced in first pass times in trials without regressions out from the region, as well as marginally reduced in go past times on the remnant, and in total times on the spill over region. The overall pattern thus suggests that readers were sensitive to information from the context, but that context was not sufficient to completely override the cost of sprouting a correlate for the remnant. The results support the claim that the processor relies on parallelism to pair a remnant and a correlate, rather than abandoning the search for a correlate and attempting to compute an entailment relation between clauses whenever possible. The alternative hypothesis according to which the apparent sprouting penalty in Experiment 1C was solely driven by accommodating a new referent is not supported by these results.
Two norming studies presented initial evidence that PP sprouting is dispreferred in focus-sensitive coordination. In the no-context completion study, participants very rarely supplied a PP as a remnant after the
In two eye tracking while reading studies, we found that PP sprouting also interferes with the on-line processing of focus-sensitive coordination sentences. In the first eye tracking study, a processing advantage for PP remnants over VP remnants was reversed in the case of PP sprouting. As long as a PP was present in the initial clause, even if the PPs evoked a different scalar dimension, the PP remnant was comparatively easy to process. Sentences in which the PP correlate and the PP were incompatible showed later delays in processing only during the first half of the experiment, after which participants apparently habituated to the mismatching scalar dimensions. In a second eye tracking study, where supportive or neutral contexts preceded target sentences with or without a correlate PP, sprouting conditions still elicited slower reading on the remnant and increased re-reading compared to conditions with an overt correlate in the matrix clause. Even though supporting contexts facilitated recovery from the PP sprouting penalty, they did not eliminate the independent penalty for violating parallelism in sprouting.
These results all support the Parallel Contrast Principle over the Scalar Advantage Principle. Whether off-line or on-line, processing focus-sensitive coordination structures is facilitated when an overt PP correlate is present in the initial clause, even though the processor must also construct an
These results relate to an important debate about the status of parallelism in sentence processing, and whether it is specific to ellipsis or conjunction or both. Prior research has routinely found that parallelism of different types eases and speeds up the processing of different types of ellipsis, including sluicing (
Focus-sensitive coordination structures, headed by conjunctions like
The construction of a scalar relation between clauses in focus-sensitive coordination is still important to the processing of these constructions, as indicated by the cost for Incompatible PP remnants observed in the first eye tracking experiment. However, we suggest that the computation of such scales is simply delayed until after a basic clausal meaning has been constructed. The importance of parallelism follows from the conceptual steps articulated in (10) above, in which the processor must locate an appropriate correlate for the remnant before an appropriate scale can be inferred. Indeed, a promising avenue of research in this vein would be to explore whether some scales are more readily accessible than others, and if so, whether they facilitate comprehension of coordination structures that require such scales during interpretation. For now, we believe that there is strong evidence that the processor prioritizes basic structure building processes when processing sentences with ellipsis, and that parallelism, which helps the processor build structure at the ellipsis site, is a particularly powerful component in recovering the intended meaning and structure, where there was once only silence.
The additional file for this article can be found as follows:
Appendices A and B contain the full set of experimental items for Experiments 1–2. DOI:
AIC = Akaike information criterion, cm = centimeters, COCA = Corpus of Contemporary American English, CP = complementizer phrase,
The type of parallelism studied in this project involves similarities between remnants and correlates, and is not the type of parallelism that relates to the syntactic and/or semantic identity condition allowing ellipsis (e.g.,
While most varieties of English license focus-sensitive coordination in the presence of explicit or implicit negation, some dialects permit a positive variant, e.g.,
Although it is conceptually possible that the processor forgoes retrieving the correlate in step 2, and simply posits a parallel structure at the ellipsis site, we think this is unlikely given evidence for similarity-based interference, characteristic of retrieval systems, from non-correlate distractor nouns in sluicing (
A reviewer suggests re-describing the entailment relation in terms of a set inclusion between properties denoted by the verb phrases. We continue to follow previous research (
Half of the items contained an additional DP in the matrix clause, e.g.,
Three of the contexts differed from the others. For items 1–3, the Supporting contexts included both the correlate and remnant PPs instead of only the remnant PP. For item 3, the Neutral context also included the remnant PP. The rest of the contexts matched the description above.
In determining whether SAP is independently motivated, a reviewer raised an interesting contrast between a structure with an
(i)
a.
Michael couldn’t study carpentry, much less chemistry.
b.
Michael couldn’t study anything, much less chemistry.
We found similar cases in COCA, with
(ii)
a.
“I couldn’t picture my Grandma as someone responsible for the death of
b.
“But the idea that Susan owed anything to
At any rate, the motivation for SAP remains conceptual, and we believe the finding that a conceptually plausible benefit for computing ready-made relations does not outweigh the general preference for parallel structures reveals the strength of parallelism biases during sentence processing.
The authors would like to thank Jack Atherton, Jenny Chim, Aura Heredia Cruz, Reuben Garcia, Angela Howard, Samantha Jew, Lexi Loessberg-Zahl, Shayna Lurya, Caitlyn Wong Pickard, Ian Rigby, and Karina Ruiz for assistance running the eye tracking experiments. Portions of this research have been presented at a UC San Diego colloquium, a UMass Psycholinguistics Workshop, and the 29th CUNY Human Sentence Processing Conference; we thank the audiences for the comments and questions, especially Chuck Clifton for suggesting Experiment 2.
The research reported in this publication was partially supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development of the National Institutes of Health under grant number R15HD072713 and an Institutional Development Award from the National Institute of General Medical Sciences of the National Institutes of Health under grant number 5P20GM103436-13. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
The authors have no competing interests to declare.