4.3 LA EDUCACIÓN EN LA ALEMANIA NAZ
4.3.1 INSTITUCIONES EDUCATIVAS
The other essential difference betw een the m odels of Golinkoff et al. (1994) and of M arkm an is the differing predictions that are m ade for certain experim ental situations by the N3C principle and the m utual exclusivity assum ption. The divergence of these predictions is based on a prem ise thought necessary for the m utual exclusivity assum ption. That is, that the child "must not generalise more than one nam e to the same thing" (M erriman & Bowman, 1989; W oodw ard & M arkm an, 1991). There are undoubtedly cases w here this occurs. H ow ever, it is im portant to rem em ber th at the m utual exclusivity assum ption is a default constraint and is probabilistic. It will only be followed, and the child thus avoid generalising a second nam e to an object for w hich it already has a nam e, w here there are no other salient factors. These other salient factors m ay not always be obvious to inspection of the experimental situation. They m ay be implicit to the child. Hoek, Ingram and Gibson (1986) studied the overextensions m ade by an infant girl through the second year of life. Am ongst the 6 distinct factors that caused early w ord extensions, w ere use of a know n w ord in preference to an unknow n w ord, and overextensions of preferred words. As a generalisation of a second nam e to an object already nam ed could be deem ed an invalid overextension in the m utual exclusivity paradigm , the factors of know n w ords versus unknow n w ords should be very carefully controlled for in any experim ents testing diverging predictions. To coin a phrase, the child does not always m ean w hat s/h e says or does. The problem is that there is no other w ay of testing w hat s/h e means.
Mervis, Golinkoff and Bertrand (1994) ran studies w ith 2- and 3-year- old subjects w ith a situ atio n w here the DLPF m odel and the d efau lt constraints m odel w ould indeed give different predictions. This situation is w here there is a novel nam e, b u t no nameless object available in an array of fam iliar objects. According to the default constraints model, the novel w ord w ould m ost probably be treated as a salient p art for which the child does not yet have a name. According to the DLPF model, argue Mervis et al., the child m ight rely on other cues w ithin the situation. For example, if the label was couched w ithin a sentence containing the phrase "this is a kind of [basic level] name", then the child m ight assume the novel label to refer to a subordinate category to w hich the know n label belonged. This is essentially non- controversial, even w ithin M arkm an's default constraints model. Cueing from other sources, such as syntactic information, will undoubtedly be used by the child to constrain the hypothesis space and m ay then override the m utual exclusivity assum ption to use the label as a second, subordinate, label for the object. If there were no sources of other cueing information obvious, then the child will use the Principle of Object Scope to assum e the novel label to be another nam e for that object. They w ould then extend the new nam e to other basic level exemplars of the object's category.
This w as indeed w hat M ervis et al. found w ith both age groups of children tested. O verw helm ingly, children appeared to assum e the novel label to be another nam e for the fam iliar object, and extend that label to an o th er ex em p lar from the sam e category. They th u s gave to the experimenter both exemplars of that category in response to the request "Can you give me the XXXX. ", followed by the request "Is there another XXXX? Are there any m ore XXXXs? ", from an array of two exemplars, their parts, and a distractor object. (XXXX here is the novel label). For a child to give the hypothesised result, s /h e m ust have given both whole object exemplars to the experimenter in response to the requests. There is a qualification to this result, however. Once the child has handed the first object in response to the first request, the following request contains the phrase "another XXXX....any more XXXXs". This is clear syntactic cueing to consider another object on the same categorical hierarchical level as the first object, that is, the second exem plar as hypothesised. Even so, the child has still selected the first familiar object in response to the novel label, w hich in itself violates the m utual exclusivity assumption.
The qualification to this is less strong, yet it is the same criticism that M ervis et al m ake of the pro ced u re of the M utual Exclusivity tests of M arkm an and W achtel (1988). The salient parts of familiar objects used in both experim ents, the parts necessarily hypothesised to be unfam iliar to the child so that they invited labelling, were in actual fact likely to be familiar to the children even if in pre-tests (which M arkm an and Wachtel adm ittedly and unforgivably neglected to carry out) they w ere unable to produce a reliable label for those parts. For example, M arkm an and W achtel used the claw of a ham m er, and Mervis et al used the handle of a shovel. This is unavoidable, however. W hole objects that are familiar to young children are only familiar in the form used if their p art is attached. In th at case, the p art will also be familiar to the children.
It is possible, too, that M ervis et al. com pounded this w eakness by having a w arm -up session w here the child was asked to hand all the parts to the experim enter after a dem onstration of w hat was a part. This was intended to indicate that a part response w as a welcome response. In actual fact, it m ay have had the effect of labelling for the child all the parts used as m em bers of the basic level category "part". N ot only then does this give the child a label for all the supposedly unlabelled part-objects, it also biases them tow ards considering categorical inclusion as a factor in the experim ental procedure. This m ight also have contributed to the child responding to the second request for an "XXXX" by extending the category for the first w hole object which it selected, and h an d in g over the second exem plar of its category available.
O ther apparently supportive research contains similar pitfalls. Blewitt (1994) dem onstrated that children aged 2- and 3-years-old w ould apply both basic level and superordinate category labels to the same objects, b u t did not ensure that sufficient salient other choices were available for the subjects.
It is probably impossible to construct an experimental protocol that can settle the debate betw een the tw o camps on this issue, w ithout raising further problems of its own. Every experiment to some extent, in that it is designed to answ er the hypotheses of the experim enter, will im plicitly reflect their intentions. The protocols of the w hole object assu m p tio n and m u tu a l exclusivity assum ptions, in rem oving all other factors bar the w hole objects and their salient parts, to some extent also force the subject's hand in selecting
the unfam iliar parts as hypothesised, w ith no other salient factors. As will be discussed in section B.2.2, this makes them particularly interesting for testing w ith WS subjects. These procedures are the only ones to exam ine the whole object a ssu m p tio n , an d by im p licatio n its c o u n te rp a rt in the DLPF fram ework, the Principle of Object Scope.
I w ill n o t discuss in any fu rth er d ep th the DLPF Principles of Reference, Extendibility and Conventionality. These are non-controversial. The first tw o m ust be followed for any subject to be able to apply a given w ord to m ore than just one object. The third is beyond the scope of this thesis. The DLPF N3C principle w ill be tested on subjects w ith WS, in a fast- m ap p in g protocol. Prior research in this area w ith WS subjects is also discussed in section B.2.2. This will also be linked to ability in exhaustive sorting in an extension to the experim ental p ro ced u re of G opnik and Meltzoff. This will be further discussed in the following section, B.1.10. In this section, previous research on the taxonomic assum ption of M arkm an and its DLPF counterpart, the Principle of Categorical Scope, will also be reviewed in depth.
B.1.10 PR IN CIPLE OF CATEGORICAL SCOPE A N D T A X O N O M IC