Parault, Schwanenflugel & Haverback (2005)

Parault, Schwanenflugel & Haverback (2005) investigated issues in early school children’s use of the similarity between a head and a modifier in deriving interpretations for novel noun–noun compounds. The subjects were divided into three groups: 30 kindergartners (mean age = 6 years 3 months, SD = 3 months, range = 5 years 9 months to 7 years 0 months), 27 second graders (mean age = 9 years 1 month, SD = 4 months, range = 8 years 7 months to 9 years 10 months), and 26 adult undergraduate college students. Another 13 adults from introductory psychology classes rated the similarity of the head and the modifier used as conceptual combination stimuli in this study.

All stimuli used in this study were sampled from compounds used by the Gagné and Shoben (1997) and Wisniewski (1996) studies of adult noun–noun compounds.

Parault, Schwanenflugel & Haverback (2005) examined the stimuli used in these two studies and selected noun–noun compounds with relatively simple vocabulary that were likely to require novel interpretations. A total of 35 conceptual combinations were used, 31 of which were the focus of the study. Another 4 nonsense combinations

20

viewed as uninterpretable by Gagné and Shoben were also included. These nonsense items served merely to provide an opportunity for all the participants to deem some items as uninterpretable. A group of 13 college students were asked to rate the conceptual combinations presented as word pairs according to the similarity of the combinations on a 7-point scale. We wanted to focus on the situation where the combination did not directly point to specific, previously known exemplars in the child’s knowledge base. In a few cases, the vocabulary used in the earlier studies was modified slightly by substituting a semantically similar but simpler head noun. This was meant only to simplify the vocabulary for the child participants. The constituent words from the targeted combinations could be found in a second-grade child language corpus (Zeno, Ivens, Millard, & Duvvuri 1995). Furthermore, a pilot study with seven children who were one year younger than the kindergartners used in this dissimilar combinations. Another four nonsense combinations viewed as uninterpretable by Gagné and Shoben were also included. These nonsense items served merely to provide an opportunity for all the participants to deem some items as uninterpretable.

The items selected for this experiment were as follows:

21

Table 2-7 Conceptual Combinations Used by Parault, Schwanenflugel &

Haverback (Parault, Schwanenflugel & Haverback 2005, p.85)

Similar Dissimilar Nonsense

Apartment hotel interpretations were tape-recorded. Each combination was presented in the following series of frame prompts.

22

(35) “What’s a __________?”

(36) “What would it be like?”

The compounds were presented to the children in random order. The entire session lasted approximately 20 minutes. The three authors coded all the participants’

interpretations independently into one of the following five types:

(i) Relational interpretations

These involved a relational linking between the two constituent objects in

which the meaning of each word was left relatively intact. In their study, an interpretation was coded as relational if the participant set up a temporary relation between the head and the modifier or set up a relation between words via a predicate phrase or a hidden link between concepts.

(ii) Property interpretations

These involved a property or part of the modifier being carried over onto the head. In their study, an interpretation was coded as property if the participant set up a

“looks like” or “made of” relation between the head and the modifier, the head was said to be “like” the modifier or vice versa, the interpretation presented parts of the modifier stuck onto the head or vice versa, or the interpretation brought an emergent meaning to mind.

(iii) Hybrid interpretations

These involved a combination of the two constituent nouns in which the properties of neither object were dominant. In their study, an interpretation was coded as hybrid if the participant’s interpretation described a conjunction of both words (e.g.,

“half head and half modifier,” “both head and modifier”), or the unique properties of the head and modifier were mentioned but properties of neither were dominant.

These three expected interpretation codes, however, did not capture all of

the interpretations the researchers received, particularly from child participants.

23

Because of this, they added two other interpretation codes to the coding scheme.

(iv) Exemplar naming

In these cases, the participants’ interpretation explicitly mentioned a non-novel exemplar that captured the combination. The following were two examples used in the task:

(37)Coat shirt: “jacket”

(38)Fork spoon: “spork”

(v)Noncombination

In these cases, the participants’ interpretation did not represent a conceptual combination of the two words. This interpretation type included instances where the participants described only one of the two terms or described both terms independently, described only attributes that both words have in common, connected each term only through “and” or a side-by-side unconnected status, stating that he or she did not know what the combination of the two constituent words would be like, or provided a completely anomalous and nonsensical interpretation for the combination.

The following were two examples used in the task:

(39)Book magazine: “a big magazine and a little book”

(40)Apple pear: “red”

After observing the results, the children seemed to be able to make both property and relational interpretations for noun–noun conceptual combinations. The children, like the adults, appeared to be sensitive to similarity as a cue in the formation of interpretations for conceptual combinations. The children, like the adults, were more likely to provide relational interpretations for dissimilar combinations than for similar ones. Furthermore, the children were more likely to provide property interpretations for similar combinations than for dissimilar ones. The children did differ from the adults, however, in that they made fewer property interpretations and failed to actually

24

combine the meanings of words more often, producing more non-combinations than the adults did. The children seemed to have some difficulty in formulating the property interpretations for similar combinations, differentially supplying noncombinational interpretations.