The broad aim of the study is to test the new proposed model to account for matemal-informant-specific variance in ratings of child behaviour problems: that representations of child, self, and own parents (and particularly own mother) wiU account for more of the variance in ratings of child behaviour than will
depression.
Specifically it will test the idea that the content and structure of representations are more closely associated with discrepancy scores between mother and criterion raters of child behaviour problems, than levels of maternal depression. That both representations and depression will account for some unique aspects of the variance in there own right, but that together they will account for more of the variance than either can alone.
These broad based aims can be further elaborated into a series of hypotheses which can be tested specifically within the design of this study:
Hypotheses relating to depression;
L Higher levels o f maternal depression are associated with higher ratings o f
child behaviour problems from both mothers* and criterion informants.
This will rephcate previous research findings that suggest that levels of maternal depression are associated with ‘real’ increased levels of disturbance in children. This will be tested by analysing the correlations between level of maternal depression and mothers, and criterion ratings of child behaviour problems.
2, Higher levels o f maternal depression are associated with greater
discrepancy between mothers* and criterion ratings o f child behaviour
problems.
This again will reflect previous research findings that led to the development of the current model used to explain discrepancy in child behaviour ratings.
Reflecting that although there may be ‘real’ differences in behaviour of children of depressed mothers, the level of maternal depression leads to perceptions of the child as being still more disturbed than others’ perceptions of them. This will be tested by analysing the correlations between maternal levels of depression and discrepancies in child behaviour problem rating.
Hypothesis relating to representations:
3, The more positive the content ofparticipants* representations o f child, self
and parents, the less discrepant child behaviour problem ratings will tend to
be.
This hypothesis forms the first part of the initial testing of the new proposed model. As has been hypothesised earlier in this section, the rating of child behaviour is associated with the closely linked representations a mother holds of her child, herself and her own mother (and to a lesser extent her father). More negative representations are expected to lead to more negative perceptions of child behaviour, and thus to more discrepant ratings. This hypothesis will be tested in a number of ways. From the repertory grid, the ratings of the elements: child, self, own mother, and own father, on each of the constructs will be
correlated with discrepancy scores, both as individual constructs and as sum total construct ratings. It is expected that the more positive each of the elements are rated the less discrepancy there will be in ratings of child behaviour problems.
From the verbal free-response descriptions, the each of the rating scales used to assess the content of the descriptions will be correlated with discrepancy scores. It is expected that the more positively the descriptions are rated on each of the dimensions the less discrepancy there will tend to be. For example, a mother whose description of herself is rated highly on the loving/unloving scale, will be expected to be less discrepant in her rating of her chüd when compared to the criterion rater.
4. The structural features ofparticipants* representations o f child, self and
parents, will be associated with discrepancy in child behaviour problem
ratings.
This forms the second part of the initial testing of the new proposed model. As representations are proposed to affect “attention, memory, and cognition” (Main
et al. 1985), it is expected that not only the content of the descriptions will differ
but that there will be a difference in the way the descriptions are given ie; they will be structurally different, and these structural differences will also be
associated with discrepancy in child behaviour ratings. For example the closely related dimensions of coherence, and substantiality, can be thought of as
measures of the degree to which an individual is able to think about the person they are describing. Those individuals who demonstrate a less sophisticated capacity to think, or reflect, characterised for example by insubstantial or incoherent descriptions, will tend to view more behaviour as problematic. Therefore, from the free-response descriptions, those whose descriptions are rated less highly on the structural dimension scales will tend to demonstrate more discrepancy in child behaviour ratings.
From the repertory grid, mothers whose representations of self, child, and own parents are judged by them to be less than ideal, will tend to be more discrepant in child behaviour ratings. Degree of fit between the current representations and the participants ideal will be measure by the distance scores from the repertory grid analysis for each of the elements and their ideal, the following distance
pairings shall be considered: child - ideal child, self as a child - ideal child, self - ideal self, self - ideal parent, own mother - ideal parent, and own father - ideal parent. Also, the greater the degree to which an individual makes distinctions between elements and constructs in their ratings of them: the intensity of the construing, the greater the discrepancy in child behaviour ratings. Intensity is opperationalised as the amount of variance explained by the first conq)onent in the principle conq)onent analysis of the grid.
Hypothesis Related to the New Proposed Model:
5. Representations and depression together better explain maternal-
informant-specific variance in child behaviour ratings than each does independently.
This will prove the culminate test of the new model and Avill be tested using multiple regression analysis. From the correlation analysis it will be possible to assess the amount of variance in discrepancy scores accounted for by measures of depression and representations. The multiple regression will allow for the