Thesis (MA)--Stellenbosch University, 2003. / ENGLISH ABSTRACT: The aim of this study was to investigate the psychometric properties of a screening test of cognitive ability, the Paper and Pencil Games Level 2 (PPG Level 2), for Tigrignaspeaking schoolchildren in Eritrea. This study represents one of the first attempts to measure cognitive ability in Eritrea. The PPG was developed in South Africa (Claassen, 1996) as a group test of general cognitive ability for children in Grade 2 and 3 (PPG Level 2), and Grade 4 and 5 (PPG Level 3). The PPG provides Total, Verbal and Nonverbal ability scores. The Verbal Scale consists of two subtests, namely (a) Verbal and Quantitative Reasoning and (b) Comprehension. The Non-verbal Scale consists of three subtests, namely (a) Figure Classification, (b) Figure Series, and (c) Pattern Completion. Although the PPG has the appearance of a standardized intelligence test, Claassen emphasized that it is best used as a screen for academic difficulties or failures. Participants were 577 Tigrigria-speaking Grade 3 children. The children were selected from schools in the capital, Asmara, from small towns, and villages. Eleven schools participated. The participants can be considered representative of the Grade 3 Tigrignaspeaking population in Eritrea. The PPG Level 2 was completed under supervision of the researcher. Classical and Rasch item analyses were conducted on the Verbal and Non-verbal Scales, respectively. The internal consistency of the Non-verbal scale can be considered satisfactory for a screening instrument (Cronbach's a = .85). Furthermore, the non-verbal items showed satisfactory fit to the Rasch model (INFIT values and OUTFIT values < 1.3 for all items), suggesting that they measure a unidimensional construct. In addition, the item difficulty estimates corresponded well with the serial order of the items, with easy items being presented earlier than more difficult items. The internal consistency of the Verbal scale was lower (Cronbach's a = .72), which can probably be attributed to the relative easiness of the items for the particular group of participants. The verbal items also showed satisfactory fit to the Rasch model. The Rasch analysis, which expresses person ability and item difficulty on the same scale, clearly showed that the Verbal items were too easy for the majority of the children. However, it should be kept in mind that the PPG is intended to discriminate among children with low ability. Hence, the observed mismatch between the abilities and item difficulties was not unexpected. It should also be noted that the serial order of the items did not correspond well with item difficulty, with some difficult items being presented early and some easy items presented late in the scale. The five subtests of the PPG were subjected to a confirmatory factor analysis. Two models were specified and compared. Model 1 specified a single general factor; this provided a more parsimonious description of the data and showed a satisfactory fit with the data, though a bit weaker than that of Model 2. Model 2 specified two correlated factors, namely a Verbal and a Non-verbal factor. Model 2 also fitted the data well, but a high correlation between the factors was observed (r = .77; r2 = .59), suggesting the presence of a general factor. The results provide support for two levels of interpretation, namely on the Total score level and the Verbal and Non-verbal level. The validity of the PPG Level 2 was further investigated by examining the correlations between the PPG scores and teacher ratings of academic achievement. Because different schools had different raters, the correlations within each of the schools were pooled to obtain an estimate of the correlations between the PPG scales and academic achievement for the total group. The pooled correlation for the PPG Total score with academic achievement was .56, for the Non-Verbal score .53, and for the Verbal score .41. The correlations for the Total and Non-verbal scores are similar to those typically reported in the educational psychology literature and provide support for the validity of these scales as a screen for academic difficulties. The results show that the PPG, which was developed in South Africa, may be fruitfully exported to Eritrea. It is recommended, however, that before the PPG Level 2 is routinely used for screening purposes with Tigrigna-speaking children, the functioning of the Verbal Scale should be re-examined and possibly some of the Verbal items should be rewritten. / AFRIKAANSE OPSOMMING: Die doel van die onderhawige studie was om die psigometriese eienskappe van 'n siftingstoets van kognitiewe vermoë, naamlik die Paper and Pencil Games Level 2 (PPG Level 2), vir Tigrigna-sprekende kinders in Ertirea te bestudeer. Hierdie studie verteenwoordig een van die eerste pogings om kognitiewe vermoë in Eritrea te meet. Die PPG is in Suid-Afrika ontwikkel (Claassen, 1996) as 'n groeptoets van kognitiewe vermoë vir kinders in Grade 2 en 3 (PPG Level 2), en Grade 4 en 5 (PPG Level 3). Die PPG lewer tellings van kognitiewe vermoë op drie vlakke, naamlik Totaal, Verbaal en Nie-Verbaal. Die Verbale skaal bestaan uit twee subskale: (a) Verbale en Kwantitatiewe redenering en (b) Begrip. Die Nie-Verbale skaal bestaan uit drie subskale, naamlik (a) Figuur Klassifisering, (b) Figuurreekse en (c) Patroonvoltooiing. Alhoewel die PPG op die oog af soos 'n konvensionele intelligensietoets lyk, beklemtoon Claassen dat dit te beste geskik is as 'n siftingsinstrument vir akademiese probleme of mislukkings. Die deelnemers was 577 Tigrigna-sprekende kinders in Graad 3. Die kinders is uit skole van die hoofstad, Asmara, klein dorpe en geselekteer. Elf skole het aan die studie deelgeneem. Daar kan aanvaar word dat die deelnemers verteenwoordigend is van die Graad 3 Tigrigna-sprekende populasie in Eritrea. Die PPG Level 2 is onder supervisie van die navorser voltooi. Klassieke en Rasch item-ontledings is op die Verbale en Nie- Verbale skale, onderskeidelik, uitgevoer. Die interne konsekwentheid van die Nie- Verbale skaal kan as bevredigend vir 'n siftingsinstrument beskou word (Cronbach se a = .85). Hierbenewens het die Nie-Verbale items 'n bevredigende passing met die Rasch model getoon (INFIT gemiddelde kwadrate en OUTFIT gemiddelde kwadrate < 1.3 vir alle items), wat daarop dui dat die items 'n essensiëel eendimensionele konstruk meet. Die moeilikheidswaardes van die Nie- Verbale items het ook sterk ooreengestem met die volgorde waarin die items in die skaal aangebied word - die maklike items is eerste aangebied en daarna die moeilike items. Die interne konsekwentheid van die Verbale skaal was laer (Cronbach se a = .72). Die laer koëffisiënt kan waarskynlik aan die relatiewe lae moeilikheidsgraad van die Verbale items toegeskryf word. Die Verbale items het egter ook 'n bevredigende passing met die Rasch model getoon. Die Rasch ontleding, wat vermoë en item moeilikheid op dieselfde skaal uitdruk, het duidelik getoon dat die Verbale items te maklik vir die meeste kinders was. Daar dien egter op gelet te word dat die PPG ontwerp is om te diskrimineer tussen kinders met relatiewe lae vermoëns. In hierdie lig gesien is die swak passing tussen vermoë en item moeilikhede nie te onverwags nie. Die volgorde waarin die items in die skaal aangebied word het ook nie goed ooreengestem met die item moeilikheidswaardes nie - sommige moeilike items is vroeg in die skaal aangebied en sommige maklike items laat in die skaal. Die vyf subtoetse van die PPG is aan 'n bevestigende faktorontleding onderwerp. Twee modelle is gespesifiseer en vergelyk. Model 1 het 'n enkele algemene faktor gespesifisieer.