Advertisement

Long-Term Representation of the Entire Alternative Set

  • Nicole Gotzner
Chapter
  • 170 Downloads
Part of the Palgrave Studies in Pragmatics, Language and Cognition book series (PSPLC)

Abstract

Previous research has shown that focus structure guides the listener’s attention leading to more detailed memory for focused elements. Experiments 1a and 1b presented in this chapter investigate whether focus particles have an impact on long-term representations of contextually provided alternatives (an earlier version of this chapter was published in Spalek et al. (J Mem Lang 70:68–84, 2014). [Chapter  3 is based on Spalek et al. (J Mem Lang 70:68–84, 2014) with some modifications. The experiments were designed by Katharina Spalek and myself and both authors wrote the article. I prepared and recorded the stimuli and analyzed all data.] Participants were presented with auditory stimuli that introduced a set of elements (‘context sentence’) and continued in three different versions: the critical sentences either contained the exclusive particle only, the additive particle even, or no particle (control condition). After being exposed to blocks of ten trials, participants were asked to recall the elements in the context sentence. The results show that both particles enhanced memory performance for the alternatives to the focused element, relative to the control condition. Thus, focus particles facilitate encoding of information-structural alternatives in memory.

Keywords

Focus particles Alternative semantics Delayed recall Long-term memory 

3.1 Introduction

The ultimate aim of language comprehension is to construct a representation of the meaning of an utterance that has been perceived (e.g., a mental model according to Johnson-Laird, 1983; van Dijk & Kintsch, 1983). This meaning representation should be long-lasting, such that an interlocutor will be able to remember the gist of a conversation at some later point in time. The present series of experiments investigates how recall of a dialog is modulated by the presence of focus-sensitive particles like only and even .

3.1.1 Focus and Alternatives in Language Comprehension

If the focus structure of a sentence successfully reflects the intention of the speaker, it necessarily has consequences for how a sentence is processed and perceived by the listener. It is well known that focused element s are somewhat privileged in language comprehension (e.g., Cutler & Fodor, 19761979; Birch & Garnsey, 1995; Birch & Rayner, 19952010; see also Cowles (2012) and Chap.  2 for an overview). Hence, focus structure guides the listener’s attention in language comprehension. Moreover, previous research indicates that focus structure seems to alter memory representations of a discourse. A number of studies have shown that focused element s are remembered better than non-focused element s and are possibly represented with more semantic detail (e.g., Fraundorf, Watson, & Benjamin, 2010; Osaka, Nishizaki, Komori, & Osaka, 2002; Sanford, Price, & Sanford, 2009; Sturt, Sanford, Stewart, & Dawydiak, 2004; Ward & Sturt, 2007).

Most previous studies on the role of focus structure in language processing implicitly relied on the assumption that the focused element in an utterance is the most important element (see, e.g., Birch & Rayner, 2010). However, theoretical semantics, in particular Rooth (1985), assumes that the function of focus is to evoke alternative expressions that can replace the focused expression. So, instead of highlighting prominence per se, alternative semantics defines focus as indicating the presence of alternatives that are relevant for interpretation (see also Rooth, 1992; Jacobs, 19831988 and Krifka, 2007 for similar proposals). We might therefore expect focus and focus particles to play a specific role in the processing and representation of alternatives. Previous research did not take into account the effect of focus structure on alternatives and therefore cannot distinguish the role of focus structure in the processing of the focused material itself from the retrieval of alternatives, and possible interactions between the two.

3.1.2 Specific Hypotheses: Lexical Meaning or Focus Association?

As outlined in the previous chapter, Fraundorf et al. (2010) have shown that contrastive focal accents enhance the representation of focus alternatives. In the experiments in this book, I compare experimental conditions that contain a focus accent and an exclusive particle, an additive particle, or no particle (as a control). Why would we expect an effect of focus particles if focus alone already introduces a set of alternatives? Whereas focus accenting indicates the presence of alternatives, the instantiation of a contextually salient set of alternatives is a necessary meaning component of focus particles . Therefore, we might expect the alternatives to become even more salient in case an utterance contains a particle compared to bare focal intonation.

As described in Chap.  2, linguistic theory divides focus particles into groups according to their specific meaning components. Exclusives like only express that the alternatives do not hold, while additives such as even and also presuppose the truth of alternative propositions. From this distinction, it might follow that the alternatives are mentally switched off if an exclusive particle is used while they become more salient in the case of inclusive particles. This hypothesis will be referred to as the lexical meaning hypothesis. The lexical meaning hypothesis predicts that exclusive and inclusive particles should differentially affect the representation of alternatives since the different groups of particles have different meaning components—exclusive or inclusive.

However, there is an alternative possibility. Since both exclusives and additives encode a set of alternatives in their semantic definition, it might be that the representation of the alternatives is enhanced overall. In other words, for both exclusives and additives we can derive the hypothesis that the speaker/hearer needs to bear in mind a set of alternatives in their mental model upon encountering a focus particle. For this reason, memory for the alternatives might be facilitated by exclusives as well as additives. I will refer to this hypothesis as the focus association hypothesis contrastive topic .1

3.2 Experiment 1a: Salience of Alternatives in Long-Term Memory

3.2.1 Goals and Predictions

Experiment 1a employs a delayed recall paradigm to investigate the impact of focus particles on long-term memory for focus alternatives. Participants were exposed to discourses that introduced sets of three elements and specified one of the elements carrying intonational focus in the critical sentences in all conditions. The manipulation was whether the critical sentences contained the exclusive particle only , the inclusive scalar particle even , or no particle as a control condition. In the test phase, participants had to recall the elements mentioned in the context sentences after a delay of about four minutes. The measure of interest was the number of recalled alternatives to the focused element (the element mentioned in the critical sentences).

According to the lexical meaning hypothesis outlined above, memory for the alternative set should be worst for exclusives, intermediate in the control condition, and best for inclusives. Since the particle only excludes the alternatives, it might be that participants forget about the alternatives more easily because they are not true of the sentence. The particle even expresses that alternatives to the focused element were also true of the sentence. Therefore, participants might recall the alternatives better if even was used.

Alternatively, the focus association hypothesis predicts that both conditions with focus operators lead to better memory for the alternative set , compared to the control condition without a focus particle due to the fact that both particles require a salient set of alternatives.

3.2.2 Method

3.2.2.1 Participants

Twenty-six native speakers of German (21 female, five male, mean age 27.1 years, SD 3.8, age range 22–31) were recruited from the subject pool at the Institute of Psychology of Humboldt University and were paid seven euros in compensation. None of them reported any vision or hearing difficulties. The data of two subjects were excluded from the analysis due to technical problems. The remaining participants were 19 women and five men with a mean age of 27.4.

3.2.2.2 Apparatus

Participants were seated in a darkened room in front of an Acer TFT monitor (type Asus 1923d) with a resolution of 1280 × 1024 and a refresh rate of 75 Hz (13.3 ms). Stimulus presentation was controlled by Neurobehavioral Systems’ Presentation software (Version 15.1). Subjects wore Sennheiser headphones with an integrated microphone. Answers in the test sessions were recorded as wav-files. The apparatus was the same in all the experiments in this book, and therefore it is only reported once in this chapter.

3.2.2.3 Materials

Participants listened to 50 pre-recorded dialogs containing two context sentences and a critical sentence. The context sentences introduced a set of three elements, a person, and an assumption about a particular event referring to the previously mentioned entities (see (1) below). The critical sentence was spoken by a different speaker and revised the assumption of the first speaker. The purpose of choosing this particular structure of the dialogs was to introduce the additive presupposition of even and to mention all elements equally often. By mentioning all elements twice, we could make sure that any differences across conditions or between the focused element s and their alternatives were not simply due to differences in mention.

The context and continuation sentences were recorded by a male speaker and the critical sentences were spoken by myself with special attention to producing similar prosodic contours across conditions. Both speakers had a middle German accent close to the standard variety of German and recording took place in a soundproof room. For the critical sentence, three versions were recorded: it contained (a) the exclusive particle nur (‘only’), (b) the inclusive particle sogar (‘even’), or (c) no focus operator as a control condition. In the experiment, one of these three versions was randomly assigned to a given experimental list (i.e., neither context sentences nor critical sentences were repeated within participants). Example (1) shows an example item.

  1. (1)

    Context sentence (speaker 1):

    In der Obstschüssel liegen Pfirsiche, Kirschen und Bananen

    ‘In the fruit bowl, there are peaches, cherries, and bananas’

    Continuation sentence (speaker 1):

    Ich wette, Carsten hat Kirschen und Bananen gegessen

    ‘I bet Carsten ate cherries and bananas’

    Critical sentences (speaker 2):

    (a) Nein, er hat nur [Pfirsiche]F gegessen

    (b) Nein, er sogar [Pfirsiche]F gegessen

    (c) Nein, er hat [Pfirsiche]F gegessen

    ‘No, he (a) only/(b) even/(c)_ ate [peaches]F

     

The three elements given in the context sentences were presented in random order (randomization was performed with a random numbers function in Excel). The second sentence was spoken with a pitch accent on the two elements (in example (1): cherries, bananas). In the critical sentences, the third element which had not been mentioned in the second context sentence appeared as the focused element (in example (1): peaches), thereby ensuring that each element was mentioned twice in the dialog. Which element of the first context sentence (first, second, or third) was focused on in the critical sentence was counterbalanced across items.

The critical sentences with particles were pronounced with a hat contour (see Féry, 1993), with a pitch accent on the focus particle as well as the focused element , and the condition without a particle had a falling pitch accent on the focused constituent. Figure 3.1 displays the mean pitch contours of the focused element for the three experimental conditions. Importantly, the accent type was the same across conditions, as can be seen from the figure.
Fig. 3.1

Mean pitch contour of the focused element in Experiment 1a

Acoustic analyses were conducted on the focused element to assess whether there were prosodic differences across conditions. Table 3.1 presents means and standard errors for duration, maximum pitch, minimum pitch, difference between maximum and minimum pitch, mean pitch, and the relative points of maximum and minimum pitch for both the stressed syllable and the entire word. The focused element s of the three conditions did not differ significantly on those parameters except for the mean pitch of the stressed syllable and the entire word and the point of pitch maximum of the entire word.
Table 3.1

Mean acoustic parameters of the focused element in the critical sentences [e.g., Pfirsiche in (1)]

 

Only

 

No particle

Even

    

Measure

Mean

SE

Mean

SE

Mean

SE

F(2, 29)

p

 

Stressed syllable

         

Duration (s)

0.24

0.01

0.23

0.01

0.23

0.01

<1

0.77

 

Max pitch (Hz)

213

2.53

225

8

207

7.63

<1

0.58

 

Min pitch (Hz)

194

2.29

197

4.38

185

3.85

<1

0.37

 

Pitch diff.

20

1.7

28

8.49

22

5.59

<1

0.58

 

Mean pitch (Hz)

203

2.18

210

3.93

195

4.53

11.58

0.0001

 

Point of max (s)

0.15

0.01

0.13

0.01

0.12

0.01

<1

0.67

 

Point of min (s)

0.16

0.02

0.16

0.02

0.18

0.02

1.52

0.23

 

Entire word

         

Duration (s)

0.49

0.02

0.49

0.02

0.49

0.01

<1

0.87

 

Max pitch (Hz)

225

8

222

8.12

212

6.66

2.5

0.85

 

Min pitch (Hz)

148

2.46

148

2.94

145

3.03

7.43

0.1

 

Pitch diff.

78

8.4

74

7.8

67

6.74

<1

0.56

 

Mean pitch (Hz)

189

1.65

186

1.74

179

2.37

7.87

0.0001

 

Point of max (s)

0.15

0.02

0.17

0.02

0.16

0.02

2.26

0.01

 

Point of min (s)

0.46

0.02

0.46

0.01

0.43

0.02

<1

0.38

 

For the construction of each of the 30 experimental items, a unique setting was chosen (e.g., the fruit bowl was used only in one item) and 90 German high-frequency nouns (more than one occurrence per million according to the DLEX database, Heister et al., 2011) from taxonomic categories served as the list of elements (see Schröder, Gemballa, Ruppin, & Wartenburger, 2012, for the category norms). In each item, there were three nouns and the number of syllables was kept as constant as possible within an item. For the second context sentence, 23 common German verbs (normalized type frequency range: 1–68) were used that described simple actions (some verbs like einkaufen (‘to buy’) were used twice) and 30 German names (male and female).

In addition to the experimental items, a set of 20 filler items was constructed in order to discourage participants from concentrating only on the list of elements. The fillers had exactly the same structure as the experimental items and the elements chosen were from various categories (either taxonomic or non-taxonomic). The critical sentences of the filler items were also recorded in three versions and participants only saw one of the conditions in the experiment.

Participants listened to ten items in a row, comprising filler and experimental items. After that, they were presented with questions about these items (see also Sect. 3.2.2.4). For the experimental items, participants were asked to recall the list of elements. The critical question always mentioned the category and the setting, for example, Which fruits were in the fruit bowl? for example (1). For the filler items, comprehension questions were presented to make sure that participants listened carefully to the entire dialog, not just to the nouns mentioned in the first context sentence. The filler questions asked about the setting, the name of the protagonist, the action, or the focus particle. A question about the focus particle was always posed in such a way that the answer could be determined easily, for example, the question Did Carsten eat bananas and cherries? requires a no-response in the context of example (1) (No, he only ate peaches).contrastive topic .2

In total, we constructed a set of 60 filler questions: 21 about the action, 19 about the setting, five about the person, and 15 about the focus particle. A participant received 20 of those questions chosen randomly from three different lists.

Each participant was exposed to all 30 experimental items and 20 filler items in five blocks of ten items containing different particle conditions. The conditions were rotated across items according to a Latin Square Design, so that one subject heard an item only once in a particular condition. This resulted in a total of three lists that were pseudo-randomized for each participant. For randomization, the program Mix (van Casteren & Davis, 2006) was used with the following constraints: no more than three filler or experimental items were presented in a row, a particular particle condition (only , even , no particle) appeared at most twice in a row, and, within a block, a category (e.g., fruit) appeared only once to control for interference effects from similar categories.

3.2.2.4 Procedure

The experiment started with an on-screen instruction informing participants about the structure of the experiment and the task they would have to perform. The instruction explained the structure of the dialogs and told the participants that they would be asked later to recall details about the stories. Subjects were told that, during the test phase, they would only have a limited time frame to respond and that they were supposed to respond aloud. After the instructions were displayed, subjects performed four practice trials (a block of four dialogs followed by four questions) and were allowed to adjust the sound volume of the headphones.

The structure of one experimental block is exemplified in Fig. 3.2. Phases of auditory discourse presentation alternated with test phases in which participants were cued for recall. In the auditory presentation phase, each trial began with the onset of a central fixation cross displayed for 500 ms. Then, a three-sentence discourse item was presented over headphones. The next trial was immediately initiated by a fixation cross. After a block of ten trials (lasting approximately four minutes), the recall phase was initiated by a self-paced button press.
Fig. 3.2

Trial sequence: delayed recall

In the recall phase, a trial again started with a fixation cross shown for 500 ms, and then the question appeared on the screen for 3 s, followed by an asterisk that was displayed for 650 ms. Subjects were supposed to respond orally as soon as the asterisk appeared on the screen. Responses were recorded, and participants had 14 s to respond. In order not to lose information if a participant responded too early (i.e., while the question was still on the screen), recordings actually started from the onset of the question and lasted 17 s. Immediately after recording, the subsequent trial was initiated by a fixation cross and the next question was displayed on the screen. All items were tested in the same order as presented during the presentation phase. Thus, the amount of delay between presentation and test was kept constant and subjects could easily keep track of the sequence.

At the end of a block (i.e., auditory presentation phase and recall phase; see Fig. 3.2), subjects were asked to perform a simple n-backward counting task progressing from two to six increments and to take a small break. This was done to reduce interference effects between blocks, because some categories were used more than once (but only once within a given block). The experiment had five experimental blocks. After the experiment, a questionnaire was administered asking the participant for basic demographic information, what he or she thought the experiment was about, and whether he or she employed any specific strategies. An entire testing session lasted about 40 minutes.

3.2.3 Results

The recorded answers were transcribed and the number of correct responses was calculated. If a subject mentioned a variant of the presented nouns (e.g., handbag instead of the presented noun bag), we coded it as a correct response. Since we were interested specifically in memory for the alternative set , we split the data into recall of the alternatives (e.g., bananas and cherries in (1)) and recall of the focused element (e.g., peaches in (1)). First, we analyze the effect of the focus particle on the number of correctly recalled alternatives. The data are presented in the left-hand column of Fig. 3.3. Without a focus particle, 70.7% of the alternative items were correctly recalled, and with a focus particle, recall performance improved to 76.4% and 77.4% correctly recalled items for only and even , respectively.
Fig. 3.3

Mean percentage of recalled alternatives (left) and focused element (right) (Exp. 1a)

We fitted a series of linear mixed effects models using the package lme4 in R (Pinheiro, Bates, DebRoy, & Sarkar, 2014) and we followed the procedure described in Baayen (2008). We started out with a model that had a single random factor (subjects) and subsequently added additional random factors and random slopes. Model comparisons by means of F tests were performed on log-likelihood values to single out the model with the best fit. Only factors that increased the model’s prediction were kept in the final model.

Alternatives. The model with the best fit included particle condition and trial number (centered) as fixed factors, subjects and items as random factors, and random slopes for trial number. The control condition was chosen as reference level, and p-values were extracted with Markov chain Monte Carlo (MCMC) sampling with 10,000 runs. A post-hoc Tukey test on the final model (with the R-package multcomp, Bretz, Hothorn, & Westfall, 2010) was carried out for the comparison of only and even . A summary of the model is given in Table 3.2.
Table 3.2

Results of mixed effects model for correctly recalled alternatives in Experiment 1a (n = 713, log-likelihood = −3506) including fixed effect estimates (top) and variance estimates (bottom)

 

Estimate

SE

t

pMCMC

 

Intercept

70.57

4.1

17.2

  

Only

6.29

2.9

2.2

0.02

 

Even

6.6

2.9

2.3

0.03

 

Trial (centered)

0.35

0.1

3.2

0.01

 

Random effect

s 2

    

Participant

230.37

    

Item

94.53

    

Random slope

s 2

    

Trial

0.13

    

The analysis revealed a significant difference between the control condition and even (t = 2. 3, SE = 2. 9, p < 0. 05) and a significant effect for only (t = 2. 2, SE = 2. 9, p < 0. 05). The post-hoc Tukey test showed that the recall performance for only and even did not differ significantly (p = 0. 90).

Focused Element. Second, we analyzed the effect of the focus particle on the number of correctly recalled focused element s. Figure 3.3 shows the mean percentages of focused element s in the right-hand columns. The focused element was correctly recalled 81% of the time without a focus particle, 82% with only , and 83% with even . Because the dependent variable was binomial (the focused element was either recalled or it was not recalled), we carried out a logit mixed model with particle condition as fixed factor and subjects and items as random factors (see Jäger, 2008). The effects of the two particles were not significant (control vs. even : p = 0. 86; control vs. only : p = 0. 62; only vs. even : p = 0. 79).

3.2.4 Discussion

The first experiment investigated the impact of focus particles on long-term memory and used dialog structures in three versions: a condition with the exclusive particle only , the inclusive particle even , and a control condition without a particle that also carried a pitch accent on the focused element . We found that both particles improved participants’ memory for the alternatives to the focused element . From the semantics of focus particles , we had predicted that focus alternatives need to become salient to the listener during the comprehension process, and have outlined two hypotheses concerning this specific process.

The results showed that it was the presence or absence of a focus-sensitive particle that affected the recall performance for the alternatives, whereas the additional meaning component (inclusive vs. exclusive) of the particle had no effect. That is, memory performance for the alternatives was not differentially affected by the fact that alternatives were included or excluded, but only by the fact that focus-sensitive particles grammatically depend on an alternative set , thereby enhancing memory for focus alternatives. This finding is in line with the focus association hypothesis outlined at the start of this chapter.

The focused element was recalled equally often in all conditions and its mean recall was quite high (about 82%). However, if we were to claim that focus particles only affected recall of focus alternatives but not of the focused element itself, we would have to show an interaction between the type of recalled element and the particle condition. We tested this differential effect in a combined analysis. We ran an omnibus generalized mixed model that contained all data coded binomially. The final model included particle condition and type of element (focused or alternative) as well as an interaction of these two factors (and it further contained trial, item, and subjects as factors, as well as random slopes for trial). The condition without a particle in the alternatives was chosen as reference level. The model showed a main effect of type of recalled element (p < 0. 0001), showing that a focused element was recalled better than an alternative. This finding indicates that focused element s gain a privileged representation (among salient alternatives). Note that all elements were mentioned the same number of times. The fact that the focused element was best remembered overall corroborates classic findings from the literature presented in Sect.  2.2.1.2

The omnibus analysis also showed significant effects of the focus particles (control vs. only : p < 0. 05, control vs. even : p < 0. 01). However, the interaction between focus particles and type of recalled element was not significant, suggesting that the effect was equally present for focused element s and alternatives.contrastive topic [4]. Yet when setting the focused element as reference level, the effect of the particles disappears, suggesting that focus particles affect foremost the alternatives.

This lack of a significant interaction was probably due to imbalances in the two data sets (e.g., there were two alternatives but only one focused item). But in any case, no strong claims about the effect of focus particles on recall of the focused element can be made. It might be the case that a beneficial effect of the presence of a focus particle is also observed for recall of the focused element itself. Importantly, though, this does not dispute the observation that the presence of focus particles improves memory for focus alternatives.

In this experiment, we did not manipulate the presence or absence of (contrastive) focus. Rather, we investigated the effect of expressions that associate with a focused element . While the study of Fraundorf et al. (2010) has demonstrated that the encoding of alternatives is improved when a contrastive accent is used compared to a non-contrastive accent, we show that the presence of focus-sensitive particles further improves the memory representation of the alternative set .

Why should focus-sensitive particles have an effect on the mental representation of focus alternatives above and beyond intonational focus marking? If we follow Rooth (19851992) and Krifka (2007), the primary function of focus is to indicate the presence of interpretation-relevant alternatives. Beaver and Clark (2008) distinguish between types of focus sensitivity that result from pragmatic forces and those that are lexically encoded. They argue that the focus sensitivity of focus particles is conventionalized, that is, lexically encoded. Therefore, the presence of an alternative set would be indicated from a combination of sources—focus itself signals its presence, and this effect is reinforced by the conventionalized focus sensitivity of focus particles . We suggest that this double signaling of the relevance of an alternative set improves the encoding of the alternatives in the mental representation of the discourse, leading to better performance in the delayed recall task.

We had tried to find a context that allowed the use of minimal pairs, that is, critical sentences which differed only in the presence/absence of a focus particle and the specific particle used. In particular, the pitch accent on the focused element also had to be identical in all three contexts. Given that the two particles we used lead to different contextual assumptions, it was not easy to find a type of context that supported both of them and the absence of a particle equally well. In the materials used, the critical sentences were a correction of previously introduced expectations. In hindsight, there are two potential problems with these items, one theoretical, one based on the exact materials used. First, because a correction was used, the focus in all critical sentences was contrastive in a narrow sense (e.g., some theories reserve the notion of contrastive focus to corrections and parallel structures; see Krifka, 2007). This led to the fact that there was a strong exhaustiveness implicature in the condition without a focus operator. That is, if we consider the sentence No, Carsten ate pears, the natural interpretation is that he did not eat the other fruits mentioned before. Hence, it could be argued that the condition without a focus operator did not differ much conceptually from the condition with only . Note, however, that the condition without a particle and the condition with only still differ theoretically, since only modifies the assertion of the sentence, while exhaustivity comes as an implicature in the case of the condition without a particle.

Second, in the items used, the use of even appeared awkward to some, but not all listeners. A rating study, presented in Sect. 3.4, confirms this intuition and this issue will be further discussed in that section. In Experiment 1b, we wanted to test whether the effects generalize to contexts where the exhaustiveness implicature is less prominent, where only and even were equally appropriate contextually, and with stimuli that have a narrative structure without a correction.

3.3 Experiment 1b: Replication with Narrative Item Structure

3.3.1 Goals and Predictions

Experiment 1b served to replicate the findings of Experiment 1a and to see whether the effects of focus particles generalize over different linguistic structures. For example, it might be that the correction also made the alternatives more salient. Since we are interested in the effect of the semantics of focus particles , we used a narrative item structure in Experiment 1b that did not involve the correction. If the effects we observed in Experiment 1a are due to the presence of focus particles , we should observe comparable effects in a different linguistic environment.

3.3.2 Methods

3.3.2.1 Participants

A total of 33 native speakers of German (21 female and 12 male, mean age 25.7 years, SD 2.65, age range 21–32) were recruited from the same subject pool and were paid seven euros in compensation. None of them reported any vision or hearing difficulties.

3.3.2.2 Materials

We created a set of 45 experimental items (30 items were based on the material in Experiment 1a and 15 additional experimental items) and 35 filler items (20 items based on Experiment 1a and 15 additional filler items). The narrative structure used in Experiment 1b followed the schema in (2). The first context sentence mentioned a set of three elements and connected the person with the setting. The second continuation sentence continued the story and served to ease accommodation. Finally, the critical sentence referred back to the list and focused on one of the elements. We again used the three particle conditions as in Experiment 1a: the exclusive particle only , the inclusive particle even , and the control condition without a focus particle. All sentences were recorded by myself.

  1. (2)

    Context sentence: Matthias erhält ein Paket mit Hemden, Hosen und Jacken

    ‘Matthias receives a package with shirts, trousers, and jackets.’

    Continuation sentence:

    Er guckte, was ihm gefiel

    ‘He considered what he liked’

    Critical sentence:

    (a) Er hat nur die [Hemden]F behalten

    (b) Er hat sogar die [Hemden]F behalten

    (c) Er hat die [Hemden]F behalten

    ‘He kept (a) only/ (b) even/ (c) _ the [shirts]F

     

Care was taken that, across items, the focused element in the critical sentence was equally often the first, second, or third element from the first context sentence. Again, the critical sentences with particles were pronounced with a hat contour (Féry, 1993) and the condition without particles had a falling pitch accent on the focused constituent.

We again conducted acoustic analyses of the focused element across the three particle conditions. Figure 3.4 displays the mean pitch contour of the focused word (e.g., Hemden in (2)) across conditions. As the figure shows, the pitch contours closely resembled each other across conditions. In comparison to Experiment 1a, the pitch accent was less prominent, which is likely due to the (lack of) correction. Table 3.3 summarizes the results of the statistical analyses. The only significant difference across conditions was the lower minimum pitch of the entire word in the condition with even .
Fig. 3.4

Mean pitch contour of the focused element in Experiment 1b

Table 3.3

Mean acoustic parameters of the focused element in the critical sentences (e.g., Hemden in (2))

 

Only

 

No particle

Even

    

Measure

Mean

SE

Mean

SE

Mean

SE

F(2, 44)

p

 

Stressed syllable

         

Duration (s)

0.24

0.01

0.23

0.01

0.23

0.01

1

0.37

 

Max pitch (Hz)

224

12.5

231

15.3

230

15.8

<1

0.78

 

Min pitch (Hz)

180

3.48

175

3.74

171

3.11

1.83

0.17

 

Pitch diff.

45

12.7

56

15.1

59

15.9

1.17

0.31

 

Mean pitch (Hz)

195

3.87

196

5.05

194

5.07

<1

0.69

 

Point of max (s)

0.1

0.01

0.11

0.01

0.1

0.01

<1

0.98

 

Point of min (s)

0.17

0.02

0.16

0.02

0.15

0.02

<1

0.5

 

Entire word

         

Duration (s)

0.43

0.01

0.43

0.01

0.42

0.01

1.85

0.16

 

Max pitch (Hz)

233

14.3

240

16.5

239

17.1

<1

0.77

 

Min pitch (Hz)

148

3.24

148

3.28

143

3.87

12.17

0

 

Pitch diff.

85

14.4

92

16.9

95

17.1

1

0.37

 

Mean pitch (Hz)

183

2.31

184

3.21

181

3.01

1

0.84

 

Point of max (s)

0.13

0.02

0.12

0.02

0.12

0.02

1

0.36

 

Point of min (s)

0.4

0.02

0.38

0.02

0.38

0.02

1

0.36

 

The items were constructed according to the same criteria as in Experiment 1a Half of the experimental items contained the definite determiner (German die or der) before the noun and half of them did not, depending on what sounded natural in the given contexts. This time, the filler items were only recorded in the version without particles and consequently there were no filler questions about the particles. We had omitted filler questions about particles in order to avoid drawing attention to their existence in the stimuli. Each participant received the same filler items with the same questions. Filler questions could concern the setting, the name of the protagonist, or the action. The experimental questions mentioned the category and the setting (e.g., Which pieces of clothes were in the package? (2)), requiring participants to recall the focused element and the alternatives.

Each participant was exposed to all 45 experimental items and 35 filler items in eight blocks of ten stimuli. The conditions were rotated across the experimental stimuli according to a Latin Square Design. This resulted in a total of three lists that were pseudo-randomized for each participant. The following constraints were set for randomization: no more than three filler or experimental trials were presented in a row, the particular particle condition appeared at most three times in a row, and, within a block, a category (e.g., fruit) appeared only once.

3.3.2.3 Procedure

The procedure was the same as in Experiment 1a, with one exception: Instead of five blocks, eight blocks were presented in Experiment 1b, and the experiment lasted 50 minutes.

3.3.3 Results

The recorded answers were transcribed and coded in the same way as in Experiment 1a. Again, we first looked at the correctly recalled focus alternatives. Figure 3.5 depicts participants’ performance for the alternatives (left) and focused element s (right). 59.7% of the items were recalled in the control condition, 64.3% with only , and 64.2% with even .
Fig. 3.5

Mean percentage of recalled alternatives (left) and focused element (right) (Exp. 1b)

Alternatives. The same mixed effects model as in Experiment 1a was fitted to the data. A summary of the model is given in Table 3.4.
Table 3.4

Results of mixed effects model for correctly recalled alternatives in Experiment 1b (n = 1476. log-likelihood = −7373) including fixed effect estimates (top) and variance estimates (bottom)

 

Estimate

SE

t

pMCMC

 

Intercept

59.7

3.9

15.4

  

Only

4.63

2.2

2.1

0.04

 

Even

4.51

2.2

2.1

0.05

 

Trial (centered)

0.14

0.06

2.4

0.03

 

Random effect

s 2

    

Participant

314.02

    

Item

141

    

Random slope

s 2

    

Trial

0.06

    

The model yielded a significant effect for even (t = 2. 1, SE = 2. 2, p < 0. 05) and only (t = 2. 1, SE = 2. 2, p < 0. 05). A post-hoc test revealed no difference between even and only (p = 1. 0).

Focused Element. The focused element was correctly recalled 75% of the time without a focus particle, 77% with only , and 76% with even . The recall of the focused element was not affected by the particle condition (control vs. even : p = 0. 59, control vs. only : p = 0. 28, only vs. even : p = 0. 85).

3.3.4 Discussion

Experiment 1b replicates the main results from Experiment 1a: The presence of a focus particle improved memory for focus alternatives (but not the memory for the focused element itself). Recall was affected by the presence of a focus particle, but not by the particular focus particle used, supporting the focus association hypothesis and refuting the lexical meaning hypothesis. We did not observe a difference for inclusive and exclusive particles. That does not mean that these particles do not affect how a listener interprets a sentence. For example, in studies on sentence processing it was found that only and even elicit different expectations concerning the upcoming material (Filik et al., 2009; Kim, 2012). However, both types of particles require a salient set of alternatives, and this seems to be what makes listeners better remember the alternatives in the long run.

The items used in Experiment 1b did not involve a corrective statement in the critical sentence, but the particles were embedded in a declarative statement. The pattern of results looked exactly the same as in Experiment 1a, suggesting that the observed effects can be traced back to the semantics of the focus particles .

Again, we carried out an omnibus analysis which showed a main effect of focus (p < 0. 0001, focused element s were again recalled better than alternatives), better recall in the conditions with only and even compared to the control condition (p < 0. 05 respectively), but no interaction between type of element and particle condition. Recall performance was lower on average in this experiment compared to the first, especially for the alternatives. We believe that this is due to two factors. First, Experiment 1b was longer than Experiment 1a (eight blocks with ten items each instead of five blocks with ten items each), which might have decreased performance due to fatigue but also more interference from previously recalled information in Experiment 1b. Second, while both the focused element and the alternatives were named twice in Experiment 1a, the alternatives were only named once in Experiment 1b (whereas the focused element was still named twice). The first factor affects recall of alternatives and the focused element equally, while the second factor only affects recall of the alternatives, which might explain why performance for alternatives dropped more than performance for the focused element .

Overall, the results of Experiments 1a and 1b extend the findings of Fraundorf et al. (2010) to the domain of focus particles . To conclude, focus particles lead to deeper encoding of the alternative set due to the fact that such particles require a salient set of alternatives.

3.4 Rating Study: Acceptability of Experimental Items

In the two experiments just presented, we attempted to create a context that was equally suitable to use the particle even and only . As described in Chap.  2, the particle even carries a presupposition of likelihood and an additive presupposition which need to be satisfied in the context. The particle only does not trigger the same kinds of presuppositions, and therefore it was difficult to find a context that was equally appropriate for the two particles.

In the post-experimental questionnaire, participants seemed to have understood the targeted meaning of the sentences. However, to some participants the sentence with even sounded somewhat awkward, especially in the first experiment where even was embedded in the correction. We therefore carried out a rating study with the items of the two experiments to quantify their contextual appropriateness. In this study, we presented the target sentences to a group of participants and asked them to rate the contextual appropriateness of the critical sentence, given the two context sentences (a scale from 1 to 7 was used, with 1 representing completely inappropriate and 7 very appropriate).

3.4.1 Methods

3.4.1.1 Participants

Twenty-four participants (six male, 18 female) from the same population as in the experiments reported in the body of the paper participated in the rating study. Their mean age was 24.5, with a range from 20 to 30. They were paid five euros in compensation. None of them had taken part in the recall experiments.

3.4.1.2 Materials

The experimental items from Experiment 1a (30) and Experiment 1b (45) were used. Three lists were created such that an item appeared only in one condition (with only , with even , control without a particle) within a given list but so that the three conditions appeared equally often on each list. Thirty-eight filler items were created that contained violations of semantic or pragmatic appropriateness. (3)-(a) shows an example where the continuation is contextually incongruous. In example (3)-(b), the use of only in combination with the affirmative yes is infelicitous. Each list contained the same filler items, such that a given list contained 113 items.

  1. (3)
    1. a.

      The adult education center offers classes in drawing, photography, and sewing

      I bet Marius has attended classes for photography and sewing

      Yes, he has painted pictures

       
    2. b.

      On the playground, there are monkey bars, swings, and slides

      I bet Sophie has used slides and swings

      Yes, she has only used monkey bars

       
     

3.4.1.3 Procedure

Participants listened to the sentences via headphones and were instructed to indicate the contextual appropriateness of the third sentence, given the preceding two sentences. Participants were required to make their judgment on a scale from 1 to 7, with 1 representing completely inappropriate and 7 very appropriate. Answers were given on a keyboard by pressing a number from 1 to 7. Each participant was presented with one experimental list only. The entire test session lasted about 25 minutes.

3.4.2 Results

Table 3.5 shows the mean ratings across conditions for the two experiments. As can be seen from the table, the sentences with even were less acceptable than the other conditions, though still far more appropriate than the filler sentences.
Table 3.5

Mean appropriateness ratings for items used in Experiments 1a and 1b

Exp./condition

Only

No particle

Even

Filler

 

Experiment 1a

6.5

6.4

5.1

1.7

 

Experiment 1b

5.6

6.0

5.1

2.4

 

We carried out a linear mixed effects analysis with all data. The final model contained the fixed factors experiment (1a vs. 1b) and particle condition (only , even , no particle) and the random factors participants and items. This time, only (Experiment 1) was chosen as the reference level since we were interested in the differences between the two conditions with particles.

The model showed a significant effect of Experiment (p < 0. 0001), a significant difference between only and even (p < 0. 0001), and a significant interaction between experiment and particle condition (only vs. no particle: p < 0. 001; only vs. even : p < 0. 0001), illustrating that the difference between only and even was smaller in Experiment 1b than in Experiment 1a. Table 3.6 shows the results of the mixed model.
Table 3.6

Results of mixed effects model for appropriateness ratings in Rating Study 1 (n = 2250, log-likelihood = −3976) including estimates, confidence intervals and p-values

 

Estimate

Lower

Upper

pMCMC

 

Intercept

6.51

6.28

6.73

0.0001

 

No particle (Exp. 1)

-0.07

-0.29

0.16

0.51

 

Even

-1.45

-1.67

-1.24

0.0001

 

Exp. 2

-0.87

-1.07

-0.65

0.0001

 

No particle:Exp. 2

0.47

0.19

0.75

0.001

 

Even:Exp. 2

0.91

0.62

1.18

0.0001

 

3.4.3 Correlational Analyses: Acceptability and Recall

In order to make sure that the reduced acceptability of the sentences did not affect recall performance, correlational analyses were run. In particular, it might be the case that the reduced acceptability of the condition with even led to worse recall, possibly masking any differences between the condition with even and only . We carried out separate correlations for the materials of Experiments 1a and Experiment 1b and separate correlations for the two focus particles , paying particular attention to the condition with even , because this is where an appropriateness effect should emerge. In Experiment 1a, the correlation between the ratings and the recall performance was r = −0. 09, p = 0. 66 for items with even and r = −0. 43, p < 0. 05 for items with only . In Experiment 1b, the correlation was r = −0. 0004, p = 0. 99 for items with even and r = 0. 08, p = 0. 62 for items with only . Hence, the only significant correlation showed that recall was worse for items which were contextually more appropriate, and this was only the case for one set of items in the presence of only .3 These observations speak against a trade-off of the effect of focus particles and contextual appropriateness, which might have masked an additional effect of lexical meaning.

The correlational analyses indicate that the slightly reduced appropriateness did not seem to have affected participants’ recall performance. Another aspect that speaks against such a trade-off is that we employed the same manipulation in two different linguistic contexts and found the same pattern of results.

Finally, the comparison to the filler items with mild pragmatic or semantic violations indicates that the sentences with even were not completely odd (5.1 out of 7 for even and 2.05 on average for filler items).

3.5 Chapter Summary

The first two experiments in this book lend support to the focus association hypothesis formulated in the beginning of this chapter. We found that the exclusive particle only and the inclusive particle even facilitated recall of contextually introduced alternatives to a similar extent. The results hence provide evidence that the presence (vs. absence) of a particle leads to better encoding of the alternative set and consequently better retrieval of focus alternatives. In addition to the effect of focus particles , we found an overall effect of focus such that the element in focus was better remembered than the alternatives. This suggests that focal elements have a privileged representation in the listener’s mind.

Footnotes

  1. 1.

    In Spalek, Gotzner, and Wartenburger (2014), we called this hypothesis the contrast hypothesis. However, since I do not assume that focused element s genuinely contrast with the alternatives (in the sense that the alternatives have to be negated), the term was relabeled here (see the discussion in Chap.  2), which predicts similar effects for both particles.

  2. 2.

    Note that in the case of a sentence without a focus particle this question is possibly indeterminate, since this is not explicitly stated by the sentence. We avoided these kinds of questions in order not to confuse participants. An example question about the particle even is Did Carsten eat several kinds of fruits?, requiring a yes response.

  3. 3.

    Note also that in this experiment (1a) ratings were similar for only and the control condition.

References

  1. Baayen, H. (2008). Analyzing linguistic data: A practical introduction to Statistics using R. Cambridge: Cambridge University Press.CrossRefGoogle Scholar
  2. Beaver, D., & Clark, B. (2008). Sense and sensitivity: How focus determines meaning. Oxford: Blackwell.CrossRefGoogle Scholar
  3. Birch, S., & Garnsey, S. (1995). The effect of focus on memory for words in sentences. Journal of Memory & Language, 34, 232–267.CrossRefGoogle Scholar
  4. Birch, S., & Rayner, K. (1995). Linguistic focus affects eye movements during reading. Memory & Cognition, 25, 653–660.CrossRefGoogle Scholar
  5. Birch, S., & Rayner, K. (2010). Effects of syntactic prominence on eye movements during reading. Memory & Cognition, 38, 740–754.CrossRefGoogle Scholar
  6. Bretz, F., Hothorn, T., & Westfall, P. (2010). Multiple comparisons using R. Boca Raton: CRC Press.CrossRefGoogle Scholar
  7. Cowles, H. W. (2012). The psychology of information structure. In M. Krifka & R. Musan (Eds.), The expression of information structure (Vol. 5, pp. 287–318). Walter de Gruyter.Google Scholar
  8. Cutler, A., & Fodor, J. (1976). Phoneme-monitoring reaction times as a function of preceding intonation contour. Perceptual Psychophysiology, 20, 56–60.Google Scholar
  9. Cutler, A., & Fodor, J. (1979). Semantic focus and sentence comprehension. Cognition, 7, 49–59.CrossRefGoogle Scholar
  10. Féry, C. (1993). German intontational patterns. Walter de Gruyter.CrossRefGoogle Scholar
  11. Filik, R., Patterson, K., & Liversedge, S. (2009). The influence of ‘only’ and ‘even’ on online semantic interpretation. Psychonomic Bulletin & Review, 16, 678–683CrossRefGoogle Scholar
  12. Fraundorf, S., Watson, D., & Benjamin, A. (2010). Recognition memory reveals just how contrastive contrastive accenting really is. Journal of Memory & Language, 63, 367–386.CrossRefGoogle Scholar
  13. Heister, J., Würzner, K. M., Bubenzer, J., Pohl, E., Hanneforth, T., Geyken, A., et al. (2011). Dlexdb - eine lexikalische Datenbank für die psychologische und linguistische Forschung. Psychologische Rundschau, 62, 10–20.CrossRefGoogle Scholar
  14. Jacobs, J. (1983). Fokus und Skalen: Zur Syntax und Semantik der Gradpartikeln im Deutschen. Tübingen: Niemeyer.CrossRefGoogle Scholar
  15. Jacobs, J. (1988). Fokus–Hintergrund–Gliederung und Grammatik. In H. Altman (Ed.), Intonationsforschungen (pp. 183–216). Tübingen: Niemeyer.Google Scholar
  16. Jäger, T. (2008). Categorical data analysis: Away from ANOVAs (transformation or not) and towards logit mixed models. Journal of Memory & Language, 59, 434–446.CrossRefGoogle Scholar
  17. Johnson-Laird, P. N. (1983). Mental models: Towards a cognitive science of language, inference, and consciousness (Vol. 6). Harvard: Harvard University Press.Google Scholar
  18. Kim, C. (2012). Generating Alternatives: Interpreting Focus in Discourse. Ph.D. thesis, University of Rochester.Google Scholar
  19. Krifka, M. (2007). Basic notions of information structure. In C. Féry, G. Fanselow, & M. Krifka (Eds.), The notions of information structure. Interdisciplinary Studies on Information Structure (ISIS) (Vol. 6, pp. 13–55). Potsdam.Google Scholar
  20. Osaka, M., Nishizaki, Y., Komori, M., & Osaka, N. (2002). Effect of focus on verbal working memory: Critical role of the focus word in reading. Memory & Cognition, 30, 562–571.CrossRefGoogle Scholar
  21. Pinheiro, J., Bates, D., DebRoy, S., & Sarkar, D. (2014). R core team (2014) nlme: linear and nonlinear mixed effects models. r package version 3.1-117. See http://CRAN.R-project.org/package=nlme\/
  22. Rooth, M. (1985). Association with Focus. Ph.D. thesis, Massachussets Institute of Technology.Google Scholar
  23. Rooth, M. (1992). A theory of focus interpretation. Natural Language Semantics, 1, 1–42.CrossRefGoogle Scholar
  24. Sanford, A., Price, J., & Sanford, A. (2009). Enhancement and suppression effects resulting from information structuring in sentences. Memory & Cognition, 37, 880–888.CrossRefGoogle Scholar
  25. Schröder, A., Gemballa, T., Ruppin, S., & Wartenburger, I. (2012). German norms for semantic typicality, age of acquisition, and concept familiarity. Behavior Research Method, 44, 380–394.CrossRefGoogle Scholar
  26. Spalek, K., Gotzner, N., & Wartenburger, I. (2014). Not only the apples: Focus particles influence memory for information-structural alternatives. Journal of Memory & Language, 70, 68–84.CrossRefGoogle Scholar
  27. Sturt, P., Sanford, A., Stewart, A., & Dawydiak, E. (2004). Linguistic focus and good-enough representations: An application of the change-detection paradigm. Psychonomic Bulletin & Review, 11, 882–888.CrossRefGoogle Scholar
  28. van Casteren, M., & Davis, M. H. (2006). Mix, a program for pseudorandomization. Behavior Research Methods, 38, 584–589.CrossRefGoogle Scholar
  29. van Dijk, T. A. & Kintsch, W., (1983). Strategies of discourse comprehension (pp. 11–12). New York: Academic.Google Scholar
  30. Ward, P., & Sturt, P. (2007). Linguistic focus and memory: An eye movement study. Memory & Cognition, 35, 73–86.CrossRefGoogle Scholar

Copyright information

© The Author(s) 2017

Authors and Affiliations

  • Nicole Gotzner
    • 1
  1. 1.Humboldt UniversityBerlinGermany

Personalised recommendations