The Dative Alternation in the Interlanguage of German Learners of English

Contents

1 Introduction

2 Phenomenon of inquiry: the English dative alternation
2.1 Basic terms and terminology
2.2 State of research
2.2.1 Semantic approaches to the dative alternation
2.2.2 Informational approaches to the dative alternation
2.2.3 Predicting the dative alternation
2.2.4 Lexical biases

3 The present study

4 Method
4.1 Material design
4.1.1 Test sets
4.2 Participants
4.2.2 American participants (group L1)
4.2.1 German participants (group L2)
4.3 Procedure
4.4 Score

5 Results

6 Discussion

7 Conclusion

References

Appendices

Appendix 1 - Test items

Appendix 2 - Questionnaire (German version)

Appendix 3 - Questionnaire (English version)

Appendix 4 - Raw data (group L1)

Appendix 5 - Raw data (group L2)

1 Introduction

Would you say you ‘...gave a good friend a scarf’ or would you rather say you ‘...gave a scarf to a good friend’? In essence, the present study is seeking an answer to this question. The grammatical phenomenon underlying the two syntactic variants above - the double object dative [a good friend] [a scarf] and the prepositional object dative [a scarf] [to a good friend] - is the so-called dative alternation. The term dative alternation captures this general ability in English to express the same event of giving with two distinct syntactic structures, as demonstrated above.

The English dative alternation has been researched extensively in the past decades. Most of the scientific attention was directed towards the meanings associated with the different syntactic variants. Gropen, Pinker, Hollander, Goldberg and Wilson (1989), Pinker (1989) and Goldberg (1995, 1997) are proponents of this approach - the semantic approach to the dative alternation.

Other researchers have tried to account for the dative alternation based on a number of informational variables (e.g. definiteness or pronominality) which were shown to exert influence on dative alternation processes in corpus studies (e.g. Thompson 1990; Collins 1995; Wasow 1997, 2002; Arnold, Wasow, Losongco & Ginstrom 2000). Recently, Bresnan, Cueni, Nikitina and Baayen (2005) combined relevant semantic and informational factors into a logistic regression model with which they are able to predict the vast majority of dative structures in a natural language corpus of more than three million words.

Additionally, a number of studies investigating the English dative alternation have revealed an interesting phenomenon commonly referred to as “lexical bias” (Wasow 2002:87). More precisely, researchers have found some dative verbs to occur more frequently in a prepositional dative variant, whereas other dative verbs seem to favor the double object variant (e.g. Davidse 1996:314f; Stallings, MacDonald & O’Seaghdha 1998:413f; Wasow 2002:87f; Arnold, Wasow, Asudeh & Alrenga 2003:60f; Bresnan & Nikitina 2003:13). Unfortunately, little is known yet as to the origin of these preferences of certain verbs to occur in particular syntactic frames.

While the English dative alternation has been widely explored for native speakers of English, as has just been demonstrated, relatively little is known about it in contexts of second language acquisition. This study aims to investigate how the dative alternation is reflected in interlanguages of German learners of English and additionally, whether German learners exhibit similar lexical biases in their choice of dative syntax as do native speakers of English.

In consideration of the language transfer issue in second language acquisition¹, it would be especially interesting to find out how German learners cope with the two syntactic variants available to them in English, as German is amongst the languages which do not show a dative alternation (Rappaport Hovav & Levin 2008:162). It seems likely that German learners of English would prefer the double object variant because it is the prevalent dative construction in German (Hameyer 1979:235).

As for the issue of lexical biases, raw data of early research on the acquisition of the English dative alternation by French native speakers (Mazurkevich 1984; Hawkins 1987) show similar lexical biases as observed for native speakers of English.² This trend suggests that lexical biases are acquired by learners of English as a foreign language. The latter will be tested empirically, by means of a questionnaire study, for German-speaking learners of English in the present thesis.

The structure of this thesis is as follows. Section 2 introduces the grammatical phenomenon of the dative alternation (henceforth shortened to DA). It elaborates on basic terms and terminology essential to DA and provides an overview of the current state of research. The chapter also outlines the two major groups of approaches to explaining the mechanisms underlying DA - semantic and informational approaches. A discussion of the phenomenon of lexical biases and the studies which reported them concludes this section. Section 3 leads to a detailed description of the aim of the present study, namely, the investigation of DA in interlanguages of German learners of English by means of a questionnaire study. The latter chapter is followed by an explication of the methodological proceedings in designing and conducting the experiment in Section 4.

The results of the experiment are presented in Section 5, which is followed by

a general discussion of the results and a brief consideration of their theoretical implications in Section 6. The general conclusion in Section 7 concludes this thesis.

2 Phenomenon of inquiry: the English dative alternation

In response to a question like the one posed at the beginning of the introduction (Would you say you ‘...gave a good friend a scarf’ or would you rather say you ‘...gave a scarf to a good friend’?), one might be tempted to say that it does not matter; possibly, because the two structures convey the same meaning. An answer like the latter would likely be based on the assumption that the two syntactic variants have the same proposition (which could be paraphrased as ‘...cause X to have Y’, where ‘X’ is the recipient, a good friend, of the theme ‘Y’, a scarf). And what is the difference then between saying ‘causing X to have Y’ and ‘causing X to have Y’? However, consider the constructions in (1) and (2):

(1) a. Ann-Marie bellowed a Christmas carol to the audience.

b. *Ann-Marie bellowed the audience a Christmas carol.

(2)³ a. *The noise gave a headache to Terry.

b. The noise gave Terry a headache.

While (1a) seems just fine, (1b) strikes as odd. In contrast, (2a) seems unacceptable, while (2b) sounds perfectly fine.⁴ If both surface structures (1a and b and 2a and b, respectively) have the same underlying proposition, why then is it that one variant is deemed acceptable in some cases, while the other is not?

Apparently, the surface structure does seem to play a role - it might even be traced back to a difference in the underlying meaning, as will be discussed later on in this thesis. For now, it is important that researchers have shown that speakers, when generating utterances involving datives like the ones above, draw on multiple variables of semantic and informational nature, which in the end determine the way the sentences come out of their mouths. In other words, it does seem to matter which one of the two syntactic variants one chooses, simply because speakers of the English language actively favor one structure over the other depending on certain variables. Most likely, this would not be the case if the two alternants were semantically identical in nature. Before the crucial variables will be inspected further (Section 2.2), the grammatical phenomenon underlying the example sentences above - the so-called dative alternation - will be treated in greater detail in the following subsection.

2.1 Basic terms and terminology

Generally, two alternative dative structures can be used to express the same event of transfer of possession in the English language:

(3) a. Cameron gave NP[a good friend] NP[a scarf].

b. Cameron gave NP[a scarf] PP[to a good friend].⁵

The first variant (3a) of the structure [V NP NP] is usually called double object dative (henceforth shortened to DOD). The second variant (3b) of the structure [V NP PP] is called prepositional dative in the literature (henceforth referred to as POD, prepositional object dative). The association between sentences of these patterns has commonly been referred to as the dative alternation.⁶

Transitive dative verbs like give generally take two object arguments in the DOD variant. One of the two objects is the direct object (henceforth Od) and the other is the indirect object (henceforth Oi). Normally, the Oi precedes the Od in linear order in that it occurs immediately postverbally (Quirk, Greenbaum, Leech and Svartvik 1994:54). The Oi typically expresses an “animate being that is passively implicated by the happening or state” denoted by the verb (ibid:741) - it expresses a recipient. The Od, on the other hand, is typically inanimate and refers to an entity directly affected by the verb, the theme (ibid:727). According to Quirk et al.’s (1994) definition, the NP [a good friend] in (3a) can be classified as Oi and thus, the recipient of the entity a scarf. The NP [a scarf] in (3a) can be identified as the Od of the verb give and thus, the theme which is being given to a recipient in this case.

In the POD variant (3b), the recipient is expressed by a prepositional phrase (PP) rather than an NP, namely [to a good friend]. Although the Oi-NP has been replaced by a prepositional object (PO) in this variant, the NP still exists in its exact same shape within the pertinent prepositional structure: PP[to NP[a good friend]. The PO is regarded as grammatically equivalent to an Oi (ibid:59).

Note, however, that the recipient NP and the theme NP have traded places in the POD’s postverbal syntagma, with the Od occurring directly after the verb and the PO in sentence-final position.⁷ The following table (Table 1) gives a short overview of the terminology employed in this thesis. Table 1 lists relevant terms and their respective abbreviations used throughout the paper. Each term is exemplified by means of a prototypical dative construction, with the constituents referred to in bold print.

Table 1 Terminology employed in this paper

illustration not visible in this excerpt

2.2 State of research

Various approaches have been put forward to explain the mechanisms that underlie DA. Among the most far-reaching are semantic and informational approaches. The class of semantic approaches mainly tries to account for dative choice in terms of verb semantics and constructional meaning, while the class of informational approaches emphasizes the role of information structure and syntactic constraints in DA (Bresnan & Nikitina 2003:2).

In a recent corpus study these two sides have been reconciled by means of logistic regression modeling. Bresnan et al. (2005) were able to predict the vast majority of dative choices in a large corpus of spontaneous speech employing a logistic regression model which factored in all variables reported to have a possible influence on dative choice. They considered no less than 14 variables, of which the majority has been shown to have an actual significant influence on DA.

It is important to note at this point that the effects of these variables could not be condensed to a single crucial variable, as had been proposed earlier by some researchers (e.g. Niv 1992; Hawkins 1994; cf. Bresnan et al. 2005:11; Wasow 2002:67). The crucial variables which play a significant role in DA will be looked into further in the following subsections (2.2.1, 2.2.3), along with an elaboration on semantic and informational approaches to DA.

2.2.1 Semantic approaches to the dative alternation

Semantic approaches to explaining DA can be divided into two main groups: the single meaning⁸ (SM) and the multiple meaning (MM) approaches, respectively (with Rappaport Hovav & Levin 2008). Proponents of the former class of approaches (e.g. Baker 1988; Larson 1988) assume both variants (DOD and POD) to be “thematic paraphrases” of one another and, consequently, to have the same syntactic deep structure. Which one of the variants will be realized in a certain situation is thought to be “case-related” (depending on information structure and other variables, but not on underlying semantics) (Levin and Rappaport Hovav 2005:130). Some SM approaches assume a transformational relation between the two alternants (e.g. Baker 1988; cf. Rappaport Hovav & Levin 2008:130f).

A substantial deficit of SM approaches, however, is that they fall short of explaining why certain types of verbs fail to alternate. This has been discussed marginally in the scenario at the beginning of Section 2. Recall sentences (1) and (2): While to bellow was deemed acceptable only in the POD construction (1a), give a headache seemed only acceptable in the DOD construction (2b) (Pinker 1989:110f).

However, if both constructions convey the same meaning, than what is the source of the limitation of some verbs (or verb classes) to just one syntactic frame?

MM approaches assume an oppositional position. According to Rappaport Hovav and Levin (2008), the MM approach is [t]he currently dominant approach (...), [as it] assumes a nonderivational relation between the variants: each is associated with its own meaning, though these are not always truth-conditionally distinguishable, and each gives rise to its own realization of arguments. (Rappaport Hovav & Levin 2008:130f)

Proponents of MM approaches (e.g. Gropen et al. 1989; Pinker 1989; Kriffka 1999, 2004; Harley 2003) thus claim different - though related - meanings, which map onto distinct surface structures (cf. Bresnan et al. 2005:4). This basic assumption is schematized in (4) showing the proposed underlying semantic representation (as advanced by Pinker 1989:82, 110) and the syntax it is associated with:

illustration not visible in this excerpt

According to MM proponents, the DOD variant (4a) is used to express a causation of possession of the recipient through the “bringing about of a ‘have’- relation” (Rappaport Hovav & Levin 2008:137), whereas the POD variant (4b) is used to express a change of place of the theme, that is, a movement of the theme to a certain spatial goal.⁹

Unlike the DOD variant, the POD variant does not necessarily entail a change of possession. However, it includes a component the DOD representation lacks; its PP contains a path constituent. The theme is thought to be transferred along this path - be it of abstract, i.e. conceptual, or of physical nature - to a goal (cf. Levin & Rappaport Hovav 2005:206; Rappaport Hovav & Levin 2008:132).

Evidence of the inclusion of a path constituent in PODs comes from the insertion of adverbials which can modify it (Tenny 1994). Compare (5a) and (5b) in this respect.

illustration not visible in this excerpt

While verb-argument combinations like those in (2) above (give idioms with fixed themes) pose a threat to SM approaches, they are often adduced as evidence for MM approaches. In fact, they are predicted by the MM-Mapping Hypothesis (4a) to categorically occur in the DOD variant (as in 6a), whereas the POD variant (6b) is judged to be ungrammatical:

illustration not visible in this excerpt

The explanation, according to Pinker (1989:110f) and other proponents of this approach, is the meaning of the idiom itself, which is assumed to be congruent with the meaning associated with the DOD variant but irreconcilable with the meaning attached to a POD structure: giving someone the creeps is causing someone to ‘possess’ the creeps through a change of state. It is not done by causing a change of place to the creeps, i.e. by moving the creeps (along a conceptual path) to an intended possessor (Bresnan et al. 2005:4; cf. Rappaport Hovav & Levin 2008:130).

On the contrary, there are other dative verbs, like the “verbs of continuous imparting of force”¹⁰ (e.g. to lower), which are claimed (according to 4b) to categorically occur in POD constructions (e.g., by Pinker 1989:110f; Levin 1993:46). Consider the following example sentences (7a and b) taken from Pinker (1989:111):

illustration not visible in this excerpt

Lowering a box to someone involves the continuous imparting of force onto it until it moves along a certain path to a goal, a ‘someone’ for this matter. It is not done by simply causing John to possess the box (which would correspond to the semantic paraphrase of 4a). Hence, to lower is predicted to occur only in the POD variant by the MM-Mapping Hypothesis (4b).

While the latter approach is plausible and well able to explain why certain verbs fail to alternate, its basic assumption of distinct meanings becomes somewhat questionable when considering a prototypical dative verb in both constructions. To exemplify this, consider the example sentences given in (8). An event of transfer of possession described with a verb like give can occur in both syntactic constructions - causation of possession (4a; 8a) or movement to a goal (4b; 8b), respectively:

illustration not visible in this excerpt

In such a case, it is very difficult to attribute distinct meanings to the sentences, which is due to their “near-paraphrase” relation (Levin & Rappaport Hovav 2005:207). What brings about this close semantic relation, according to MM proponents, is the meaning of the verb itself. Give and other members of the “transfer of possession” verb class are said to typically entail an abstract movement (Krifka 1999; cf. Levin & Rappaport Hovav 2005:207). It is this entailment of movement, as well as the fact that these verbs “lexicalize caused possession” (Rappaport Hovav & Levin 2008:135), which causes their propositions to be “truth conditionally indistinguishable” (ibid:130) according to semanticists. However, what determines which variant will be chosen when both are near-paraphrases?

A plausible account of the dative alternation like the MM approach is very appealing, but, as so often, the picture is not as simple and ultimately the DA phenomenon is more fine-grained than it appears at first sight. Bresnan and Nikitina (2003) and Bresnan et al. (2005), for example, have provided data that partially refute the MM-Mapping Hypotheses: In usage data obtained from the internet, they found instances of verbs of continuous imparting of force in DOD constructions, as well as examples of give idioms like the one above, give the creeps, in POD constructions. Sentences (9) and (10) give just two examples cited in Bresnan et al. (2005:5f):

(9) “(...) a stench or smell is diffused over the ship that would give a headache to the most athletic constitution.” www.downhomer.com/Webmag/2000/0007/page36.html

(10) “(...) Buddha lowered him the silver thread of a spider as his last chance for salvation.” http://www.inch.com/_fujimura/ImofGrmain.htm

Because both (9) and (10) are grammatically acceptable, it seems no longer reasonable to assume strict categorical differences when it comes to the object realizations of give idioms with fixed themes and, for example, verbs of continuous imparting of force, as is suggested in MM approaches. Rather, it seems that other factors than just mere semantics are at play in DA, inducing native speakers to accept sentences like (9) and (10), while still rejecting sentences like (6b) and (7b).

The next subsection deals with informational approaches to explaining DA. One goal of this section will be to discuss what kinds of factors are able to successfully compete with the strong biases of give idioms and ultimately make sentences like (9) and (10) acceptable.

2.2.2 Informational approaches to the dative alternation

Proponents of informational approaches (e.g. Halliday 1970; Givón 1984; Thompson 1990, 1995; Hawkins 1994; Collins 1995; Arnold et al. 2000; Wasow 2002) attribute the choice of dative syntax to factors such as discourse accessibility, closely related variables like definiteness and pronominality, and syntactic constraints such as the principle of end-weight (cf. Bresnan & Nikitina 2003:2).

The phenomenon of end-weight describes the predisposition of English and other languages to place longer and more complex phrasal constituents after shorter and less complex ones. The notion was introduced by Otto Behaghel for the German language as early as 1909. Thomas Wasow (e.g. 1997, 2002) and a number of other researchers (e.g. Hawkins 1994; Arnold et al. 2000) explored its scope in DA processes:

In an attempt to firstly evaluate definitions proposed for grammatical weight¹¹ and to single out the one with the highest predictive capability, Wasow (2002) conducted a corpus study.¹² His findings show that (i) the “number of words¹³ dominated” in a phrasal constituent (proposed by Hawkins 1990), (ii) the “number of nodes dominated” (proposed by Hawkins 1994), as well as (iii) the “number of phrasal nodes” (proposed by Rickford, Wasow, Mendoza-Denton & Espinoza 1995) were all three significant predictors (p <0.001) of constituent placement in English dative sentences (Wasow 2002:17); were in fact “statistically indistinguishable” (ibid:31f). Having established that a lexeme count (i) and measures of syntactic complexity (ii, iii) are equally well suited to predict constituent ordering, Wasow assumes “a single conception of weight relevant to [DA.]” (ibid:24).

The results of Wasow’s (2002) consecutive corpus study show a clear effect of constituent weight on syntactic alignment in dative sentences: In DOD variants the (sentence-final) theme NP tends to be longer, whereas in POD variants, the (sentence-final) recipient NP tends to be longer. He calculates that in both variants the final constituent is on average 3.5 times heavier than the constituent occurring immediately postverbally. According to Wasow, this is an unequivocal indication that DA processes are sensitive to the relative weight of its postverbal arguments (ibid:30).

Before moving on to other informational aspects that affect dative syntax, the scope of the principle of end-weight shall be illustrated briefly. Recall example sentences (9) and (10). Using definition (i) of end-weight it is now possible to identify the PP in (9) as heavier than the theme NP, as it contains relatively more words. The relative difference is even greater in (10), where the recipient NP contains a single pronoun and the theme NP, in comparison, is heavy, containing six words.

Although an idiom like give a headache (9) has a very high likelihood to occur in DOD constructions, when the recipient is ‘too heavy’ (i.e. when the alignment of a DOD variant would violate the principle of end-weight), it can be shifted towards the end of the sentence. This holds vice versa for POD variants (10) with an otherwise heavy theme NP in immediate postverbal position (Bresnan et al. 2005:7).

Collins (1995) emphasizes the “need for a communicative approach” examining the information structure of alternative dative constructions (Collins 1995:38f). To discuss “newness” of information, Collins (ibid:39) cites (11):

illustration not visible in this excerpt

He reasons that (11a)...

“is likely to be preferred over [(11b)] in a context where the flowers represent new information (i.e. it is established that Tom gave his aunt something and that it is news that that something was the flowers), and conversely [(11b)] is likely to be preferred if the news is the identity of the flowers-recipient.”¹⁴

(Collins 1995:39)

However, Collins notes that in general (11b) seems more natural than (11a); presumably because the definite article of the theme NP suggests that the theme has been established in previous discourse; i.e. is “discourse accessible” (ibid).

In a corpus study¹⁵ of contemporary (spoken and written) Australian English, Collins has investigated the impact of variables such as givenness of information and related variables of definiteness, pronominality, as well as end-weight on DA. He defines discourse accessibility as “the extent to which, or the ease with which, an entity is recoverable from earlier in a discourse or from the speech situation, or in some cases from both” (ibid:41).¹⁶ An entity can be classified as “given” if it is referred to in the immediate context of the speech situation; it is rated “new” when it is introduced into the speech situation for the very first time (ibid:41f).

Definiteness is defined as “information that is presented as sufficient to distinguish the referent from everything else” (ibid:45). Markers of definiteness are, for instance, determiners or proper names, as well as personal pronouns. Definiteness is thought to correlate strongly with accessibility, as definite entities are usually “identifiable on the grounds of [their] previous mention” (ibid:46). Pronominality was thought to be connected to definiteness as well as accessibility in the respect that “anaphoric pronouns typically signal that an entity is textually recoverable” (ibid:44).

Collins (1995) also takes the principle of end-weight into account as he reasons all four variables to be intricately correlated. This has already been displayed for accessibility, definiteness and pronominality. Wasow (2002) points out that reference to entities which are already established in the context can be short (be it in the form of a pronoun or not) and therefore, end-weight would also be correlating with givenness of information and definiteness and pronominality for that matter (ibid:68). Collins’ (1995) corpus study results¹⁷ are aptly summarized by Bresnan et al. (2005:8):

Collins noted that double object constructions are polarized on scales of discourse accessibility, definiteness, pronominality and length in words, with the ‘Receiver’ [recipient] having the more prominent (...) properties on these scales than the ‘Entity’ [theme].

Collins dubbed this observation the “Receiver/Entity Differentiation” (ibid:47). This difference between recipient and theme becomes evident if one regards the distribution of informational properties per construction type in his data (Figure 1).

Figure 1 Average percentage of given, definite and pronominal NPs per construction type based on Collins (1995) tabular data.

illustration not visible in this excerpt

The figure clearly shows that the recipient in DOD constructions “inherits the more prominent [informational] properties” (Bresnan et al. 2005:8) than the theme, while the difference between recipient and theme is milder in the POD constructions. Regardless of the palpable dissociation between recipient and theme properties however, the distribution of informational properties shows overall tendencies. Given and definite, as well as pronominal NPs seem to have a general tendency to occur in DOD constructions rather than POD constructions. The total of the latter properties seem to favor the immediate postverbal slot rather than the sentence-final position.

[...]

---

¹ The terms second language acquisition and foreign language acquisition are used synonymously throughout the thesis.

² Note, however, that the researchers did not interpret their data in terms of lexical biases.

³ This example is taken from Rappaport Hovav and Levin (2008:131), who make reference to Green (1974) and Oehrle (1976) for bringing it into discussion.

⁴ The ratings as ‘ungrammatical’ (*) in this paper are mostly based on Levin (2003:45-47).

⁵ There are dative verbs that take the preposition ‘for’ in the prepositional variant rather than ‘to’. In this paper, however, focus lies on dative verbs which require a prepositional complement with ‘to’ in the POD variant as this constitutes the majority of POD cases (e.g. Hawkins 1987:22).

⁶ The phenomenon has also been dubbed dative shift in the literature, but this term was discarded here for it implies a transformational association between the two variants which is controversial (e.g. Collins 1995:36).

⁷ Because there is some controversy surrounding the terms ‘direct’ and ‘indirect object’, they will be avoided if possible and substituted for the names of the thematic roles of the objects. ‘Direct objects’ will thus be referred to as themes, and ‘indirect objects’ as recipients. Note, however, that these terms will serve merely as labels and do not imply an affiliation with a particular theory of event construction.

⁸ Single meaning approaches comprise transformational and neo-transformational approaches. See Levin and Rappaport Hovav (2005:196ff) for a review of these accounts.

⁹ Note at this point, that the thematic role of the goal is compatible with that of the recipient. While in ‘Send it to Jennifer.’ the object of to, [Jennifer], can denote a spatial goal, it can likewise denote a recipient; the two thematic roles would be indistinguishable in such a case (Newman 1996:90f; Haspelmath 2003:219f; cf. Rappaport Hovav & Levin 2008:132f).

¹⁰ This term was coined by Pinker (1989) and is mostly being adopted by other researchers.

¹¹ See Wasow (2002:16f) for a comprehensive list of definitions proposed for grammatical weight.

¹² The study was done on the basis of the Brown corpus; Penn Treebank (Marcus et al. 1993).

¹³ This refers to an orthographic word.

¹⁴ This observation is concordant with Quirk et al.’s notion of “end-focus” achieved through

intonational emphasis on the final constituent (Quirk et al. 1994:1209, 1396) and, for example, Gundel’s “given before new principle” (Gundel 1988:229; cf. Wasow 2002:65).

¹⁵ For details of the corpus underlying the study see Collins (1995:36f).

¹⁶ What, e.g., Bock & Irwin 1980; Collins 1995; Prat-Sala & Branigan 2000) call “discourse accessibility” is dubbed “communicative dynamism” in Quirk et al. (1994:1396) and “newness of information” (e.g. Arnold et al. 2000) or “givenness” (e.g. Bresnan et al. 2005) by other researchers. These terms are regarded as synonymous in this paper.

¹⁷ Collins’ (1995) findings largely replicate those of an earlier study conducted by Thompson (1990).