Where then does the “mice-eater / *rats-eater” contrast come from? The general answer
is that it must somehow follow from Universal Grammar. The best specific answer available in
the early 1980s, when Gordon began his research, was that of Paul Kiparsky, who had been the
first to point out the phenomenon. Kiparsky’s (1982) account was based on Level Ordering
(therefore building on earlier work of Siegel 1979 and Allen 1978), within the framework of
Lexical Phonology. Roughly speaking, the idea was that there are three distinct, ordered levels of
word formation processes in the lexicon. Level 1 is the access-point for stored forms, including
those with irregular inflection. Compounding and regular derivational morphology (i.e.
derivational affixation that does not alter the phonological form of the base) occur at Level 2.
Regular inflectional morphology is inserted at Level 3. Thus, Level 1 can supply eat and any of
the forms mouse, mice, and rat to Level 2. There, the verb eat can be combined with the
her dominant response pattern. Gordon (1985: 85) indicates there was only a single error of the
puller-wagon type in his entire study. (The child's age is not specified.) Hence, even the
youngest group of children in his study (N = 11, age 3;02 to 4;00, 18 items per child) made fewer
such errors (at most 1; i.e. 0.5% of opportunities) than did the 3-year-olds (N = 12, age 3;00 to
3;10, 24 items per child) in Clark et al.’s (1986) study (10 errors; i.e. 3.5% of opportunities).
This may result from Gordon's holding the “_-eater” frame constant, and thereby reducing the
processing demands on the child.
Clahsen et al. (1992) successfully replicated Gordon’s study in German, where nominal
inflection is more complex than in English. For adults, possible plural markers
include -s, -en, -er, and -e. Clahsen et al. argue that adults take the “regular” plural ending to
be -s, but certain children in their study took it to be the more frequent ending -en. Either way,
each child avoided his or her “regular” ending inside compounds, and allowed the others.
derivational morpheme -er; and then any of the compounds rat-, mouse-, and mice-eater can be
constructed and passed onward. At Level 3 the stem rat could in principle be combined with the
regular plural-marker -s, but not if it is already contained inside a larger word. Hence, *rats-eater
cannot be generated.
Gordon, in his 1985 paper, was careful to point out that the theoretical importance of his
findings would remain, even if Kiparsky’s Level-Ordering story came to be replaced by another
account of the regular / irregular contrast. Indeed, the general Level Ordering thesis has since
fallen out of favor. (See Spencer 1991: 179–183 for a brief synopsis of the principal reasons.)
One challenge to the thesis comes directly from plurals-within-compounds.
On the one hand, Senghas et al. (1991) have demonstrated that adult English-speakers
judge irregular plurals to be far more acceptable than regular plurals in the modifier position of a
novel compound. Futhermore, as we have just seen, children as young as 3 years old exhibit the
same pattern in Gordon’s EP task. On the other hand, there exist a great many exceptions to this
pattern, and they belong to several different categories. Two cases were actually addressed in
Kiparsky’s original proposals. First, as discussed above, pluralia tantum include morphologically
regular plural marking, but are necessarily lexically listed and therefore must be handled at Level
1. As expected, this type of plural-marking is unproblematic within compounds.
A second type of exception concerns examples like Human Resources Department. Here
Kiparsky proposed an explanation in terms of recursion. The idea is that a morphologically
complex output of the word-formation system (e.g. the plural compound human resources)
sometimes comes back and gets listed as a single unit in the lexicon. When it does, it can be
included within another compound in the same way as a pluralia tantum. Also, as a consequence
of being lexically listed, the form can have either a more specialized meaning, or an entirely
different meaning, than its compositional semantics would indicate (cf. human resources,
Alegre and Gordon (1996) discuss a number of additional types of exception. One
concerns modification by an inherently quantificational noun. For example, a week-long seminar
can last only one week, while a weeks-long seminar must last longer than one week. Another
type of exception, which the authors term the “heterogeneous” type, includes examples like
publications catalog, where the use of plural-marking seems to indicate that the publications
listed in the catalog are heterogeneous in nature (i.e. that multiple publication-types are
represented). This intuition is brought out fairly sharply by the observation that a mineralogist,
who presumably studies many different kinds of rocks and is interested in the differences among
them, can be called a ?rocks expert much more readily than a simple pile of rocks can be called a
The main type of exception that Alegre and Gordon investigate is a case where the plural
modifier in a nominal compound crucially needs to be modified by an adjective: [new books]
shelf versus *books shelf. The authors’ analysis of this case is as follows. First, they argue that
Kiparsky’s proposal of a recursive loop is on the right track, but needs to allow full-on syntactic
objects to be stored as lexical items. This position is supported by examples like the how-can-he-
Second, they propose that the human sentence-processing system prefers not to posit the
type of structure that is required in order to have a syntactic object inside a morphological object
inside a syntactic object. Hence, faced with a choice between [
red rat] eater] (= ‘eater of
21 A crucial assumption is that this type of recursion is constrained in some way, because otherwise forms like
*rats-eater would be (or at least, would rapidly become) fully acceptable, as a result of feeding the NP rats back
into the lexicon as a stored expression.
red rats’) and [
rat eater]] (= ‘red eater of rats’], it favors the latter parse. Yet, if the
modifier bears regular plural marking, then a parse involving the compound noun *rats-eater
rats eater]] (= ‘red eater of rats’]) is strongly dispreferred, and the parse that
interleaves syntactic and morphological structure becomes the best available option: [
rats] eater] (= ‘eater of red rats’).
Alegre and Gordon argue that far from undermining the work in (Gordon 1985), the
discovery of these apparent exceptions to “Kiparsky’s Generalization” simply indicates that the
contributions of UG must be even richer than previously thought, in order to account for the
complex system (whatever it is exactly) that adults are using. Moreover, the discovery of even
finer-grained patterns in the adult data, such as the interaction of regular plural-marking with the
presence/absence of an attributive adjective, creates wonderful new opportunities to assess how
early the adult system is actually present in the child.
To this end the authors carried out a new child-language experiment, this time checking
children’s preferred interpretation of structurally complex compounds like red rat eater and red
rats eater. The method was picture-selection, where the child chose between two side-by-side
images. For example, in one image there might be a red monster eating a blue rat, while in the
other there would be a blue monster eating a red rat. The subjects were 36 children (12 each) at
As noted above, a third proposal is needed to block structures like [
rats] eater]. If these
were allowed we would have the unwanted prediction that *rats-eater is fully acceptable (or at
least as acceptable as red rats eater, given that the derivation would be essentially the same). The
authors’ proposal on this point is not entirely clear to me, but seems to involve the idea that an
unmodified nominal modifier with plural marking is necessary given the “heterogeneity”
interpretation (which would then be anomalous in the case of a rat-eater).
C# HTML5 Viewer: Deployment on ASP.NET MVC
the size of SDK package, all dlls are put into RasterEdge.DocImagSDK a Home folder under Views according to config in picture above. RasterEdge.XDoc.PDF.dll. how to copy pdf image to word; how to copy images from pdf to word
C# HTML5 Viewer: Deployment on IIS
reduce the size of SDK package, dlls are not put into Xdoc.HTML5 dll files listed below under RasterEdge.DocImagSDK/Bin directory and paste to Xdoc see picture). cut image from pdf online; paste jpg into pdf preview
the ages of 3, 4, and 5 years. Each child judged four test items. The child received either four
items with plural marking (red rats-eater), or four items without it.
The main findings were as follows. When there was no plural marker in the compound,
children in all three age groups preferred the “non-recursive” interpretation (e.g. red eater of blue
rats), as expected if children have the adult system in place very early. When the plural marker
was present, the children’s preference reversed, again as expected if children master the system
very early. At age 3, the children on average selected the “recursive” interpretation 1.0 times (out
of four opportunities) when the modifier was singular, and 3.33 times when it was plural. For 4-
year-olds the corresponding averages were 0.67 and 2.5, and for 5-year-olds they were 1.5 and
2.5. (The contrast between the singular and plural conditions was robustly significant by
ANOVA, while there was no significant effect of Age nor a significant interaction of Age with
In sum, the work of Gordon and Alegre has provided powerful evidence that English-
learning 3-year-olds already know a great deal about the morphology and syntax of synthetic
compounds—indeed, far beyond what they could realistically have inferred from their input.
Clark et al.’s well-known (1986) finding that English-learning preschoolers occasionally produce
compounds of the form puller-wagon (section 6.2.2) seems best explained in terms of task
demands, rather than erroneous decisions about the grammar of -ER compounds, given that none
of the children in the study produced the error consistently. On the other hand, Clark et al.
obtained a robust finding that when English-learning children are asked to create a novel word to
name a type of person or a type of physical instrument, younger children (3- and 4-year-olds)
mainly employ bare-stem compounds like water-person, while older children (5- and 6-year-
olds) mainly employ adult-like synthetic compounds (e.g. box-mover).
C# Raster - Modify Image Palette in C#.NET
Read: PDF Image Extract; VB.NET Write: Insert text into PDF; edit, C#.NET PDF pages extract, copy, paste, C#.NET to reduce the size of the picture, especially in copy image from pdf preview; how to cut pdf image
C# Word - Document Processing in C#.NET
Open(docFilePath); //Get the main ducument IDocument doc = document.GetDocument(); //Document clone IDocument doc0 = doc.Clone(); //Get all picture in document copying a pdf image to word; how to cut picture from pdf file
6.2.4 Synthetic Compounds in Romance Languages
Turning to languages other than English, we have seen that French (together with the other major
Romance languages) does not allow a morpheme-by-morpheme counterpart to English synthetic
compounds like dishwasher. Recall that this fact can be explained in terms of Beard’s
Generalization. If the “unmarked” position for French adjectives is post-nominal, then modifiers
in compounds should likewise follow the compound’s head, and will end up competing with the
derivational suffix -eur for the right edge of the word.
Hence, as soon as the child recognizes that the target language puts modifiers to the right
of a head noun, and that derivational morphology is suffixal, he or she should (ideally) be on the
lookout for exocentric V-N compounds like lave-vaisselle, and for phrasal compounds like
laveur de vaisselle, both of which are in principle possible in this language type. Hence we might
expect one or both of these forms to be relatively early acquisitions.
Clark (1993: ch.10) reviews studies eliciting novel agent nouns and/or instrument nouns
that had been conducted with children acquiring English, Icelandic, Hebrew, French, or Italian.
For French there were no EP data available for children younger than 5, but for Italian there was
work by Lo Duca (1990) with children ranging from 3;03 to 7;10. Clark (1993: 194–195) reports
that Lo Duca used an elicitation prompt of the form, “Com[e] si chiama quello che fa le pizze?”
(‘What do you call a person who makes pizzas?’), and that more than two-thirds of the responses
from 3-year-olds used exocentric V-N compounds. As one looks at progressively older children,
one finds increasing use of derivational suffixes, which is consistent with adult practice. (Adults
reportedly find V-N compounds more appropriate for instruments than for agents.)
Thus, Italian V-N compounding appears to be well-established by around age 3 (and
quite posssibly earlier) as a fully creative process of word-forrmation. As the child gradually
acquires the derivational morphemes that Italian provides for creating novel agent nouns, it
seems the use of V-N compounds for agents is increasingly preempted by the more specific
terms that become available. The earliness of V-N compounding, and the fact that children
initially use it even more extensively than adults do, are both consistent with the idea that
broader characteristics of the language may have “primed” the child to acquire V-N
compounding very early.
6.3 The Acquisition of Bare-Stem Endocentric Compounding
Turning now to bare-stem compounds, some of the questions that a language learner will need to
answer are the following:
Is bare-stem endocentric compounding a creative process in my target language?
If so, can it be used recursively?
Is the head on the left or the right side?
Are there any linking elements that occur inside the compounds?
If so, what determines their distribution?
Here I will focus on (2a,b). Given that Namiki’s Generalization (discussed in section 6.1.3) has
held up well over the years since it was proposed, I will also assume that creativity and
recursivity can be treated as a package.
6.3.1 Origins of The Compounding Parameter
Beginning with Snyder (1995), and in subsequent work up to the present, I have been
investigating what I refer to as “The Compounding Parameter” (TCP). In its current formulation
(e.g. Snyder 2012), TCP concerns the availability of a mechanism that is essential for, among
other things, the semantic interpretation of novel endocentric compounds. To a first
approximation, however, we might think of TCP as a simple yes/no specification of whether
bare-stem endocentric compounding is creative.
In Snyder (1995) I first proposed that certain “complex predicates,” including adjectival
resultatives (e.g. wipe the table clean) and separable-particle constructions (pull the lid off), are
possible only in [+TCP] languages. This hypothesis was suggested by work on Dutch (Neeleman
1994) and Afrikaans (LeRoux 1988) arguing that in those languages, both adjectival resultatives
and verb-particle combinations often have the morphological status of compound words. Even
though the same is not true in English, I began to explore the possibility that there is nonetheless,
even in English, a more abstract connection between complex predicates and compounding.
A comparison of Germanic languages with Romance languages suggested that creative,
bare-stem endocentric compounding (which is available in all the Germanic languages but none
of the major Romance languages) might be a relevant type of word formation, because all the
Germanic languages have adjectival resultatives and separable-particle constructions that are
comparable to the ones in English, while none of the Romance languages do. A small-scale
survey of the world’s languages lent plausibility to the idea, because for the languages sampled,
there was at least a one-way implication: Every language with adjectival resultatives and/or
separable particles had bare-stem endocentric compounding as a creative process of word
My next step was to check for a connection between complex predicates and
compounding in language acquisition. I decided to focus on the acquisition of English, because
at the time the CHILDES database (MacWhinney and Snow 1990) already included more than a
dozen longitudinal corpora for English, while for any other language there were considerably
fewer. I also decided to focus on separable particles with transitive verbs, because unlike
adjectival resultatives they are used frequently by both adults and older children; and because
For details, see Snyder (2001) and the update regarding Basque in Snyder (2012).
with transitive verbs there is often a direct object intervening between the verb and the particle,
which reduces the likelihood that a child’s verb-particle combination is simply an unanalyzed
Given the high frequency of particle constructions, and as it turned out, novel N-N
compounds, it was possible to identify a fairly precise point in each child’s corpus where the
child went from never using the given structure, to using it frequently, correctly, and with a
variety of lexical items. (In Snyder 2007 this point is referred to as the age of FRU, for “First
use, followed soon after by Regular Use with varied lexical items.”) In the case of novel
compounds, there was a worry that the frequency might be too low to see a sharp change at the
point of acquisition, but fortunately the children went through what Brown and Hanlon (1970:
33) termed a “brief infatuation”: when they discovered bare-stem endocentric compounding, they
treated it like a new toy. Hence there was no difficulty in identifying any child’s FRU.
The result was an extremely tight correlation, with a best-fit line that closely
approximated an identity function: the FRUs for compounding and particles were consistently
very close together in time, and often occurred during the same recording session. This pattern
has held up quite well as more longitudinal corpora have become available from children
acquiring English. In Snyder (2007: 92–-93) I reported an updated version of the analysis, based
on the 19 highest-quality longitudinal corpora available at that time. The ages of FRU for
compounding ranged from 1.85 to 2.59 years, and for particles ranged from 1.85 to 2.56 years.
Pearson’s r was .937, indicating that 88 percent of the variability in ages of FRU for either
compounds or particles was predicted by the ages of FRU for the other (t(17) = 11.1, p < .001).
Of course, one needs to be careful when interpreting this type of correlation, because
children go rapidly from knowing very little about their target language to knowing a great deal.
As a consequence, many different measures of language ability will show some degree of
correlation. I therefore obtained a quantitative measure of general linguistic development, and
applied the statistical technique of partial correlation. More precisely, I determined that the
children, at the point of their FRU for verb-particle constructions, had an average MLU (mean
length of utterance) of 1.919 morphemes. For each child I then determined the age at which the
MLU had first reached 1.919 morphemes.
Thus armed with a general developmental predictor of when each child would begin
using verb-particle combinations, I applied the statistical method of partial correlation. This
allowed me to see what would happen if I removed all the variability in the ages for particles and
compounds that could be explained in terms of the general developmental predictor, and just
looked for a correlation in whatever variability was left over. The correlation between the MLU-
based measure and the ages of FRU for verb-particle combinations was quite strong (r = .8690),
which indicated that the MLU measure was a good control. Nonetheless, after partial correlation,
when all the variation that could be explained by the MLU-measure had been “partialed out,”
there was still a robust correlation between compounds and particles: r
= .799, t(17) = 5.31,
p < .001. Hence, the association of particles with creative, endocentric compounding goes well
beyond what one would expect on general developmental grounds, and instead seems to be a
deeper, grammatical connection.
6.3.2 Further Tests of the Syntax–Compounding Link: Japanese and German
The proposed link of certain complex predicates to creative endocentric compounding has been
tested acquisitionally in both Japanese and German. Japanese is a [+TCP] language, with
creative bare-stem endocentric compounding as well as adjectival resultatives, but no separable-
particle construction. Japanese-learning children acquire endocentric compounding considerably
Documents you may be interested
Documents you may be interested