in a recognition test, participants typically make more hits and fewer false alarms on low-frequency...

1
In a recognition test, participants In a recognition test, participants typically make more hits and fewer false typically make more hits and fewer false alarms on low-frequency words compared to alarms on low-frequency words compared to high frequency words (A pattern of high frequency words (A pattern of results known as the mirror effect, results known as the mirror effect, Glanzer & Adams, 1985). In most of these Glanzer & Adams, 1985). In most of these experiments high and low frequency words experiments high and low frequency words are mixed at both study and test. are mixed at both study and test. Experiments 1 and 2 show that the hit- Experiments 1 and 2 show that the hit- rate component of the mirror effect is rate component of the mirror effect is eliminated when participants study a pure eliminated when participants study a pure list of either low or high frequency list of either low or high frequency list. The pattern of false alarms also list. The pattern of false alarms also changes depending on the type of list. changes depending on the type of list. We suggest that our current pattern of We suggest that our current pattern of results as well as previous results can results as well as previous results can be understood if one assumes that there be understood if one assumes that there are two-dimensions of information are two-dimensions of information available for recognition decisions: one available for recognition decisions: one that is used primarily for recognition of that is used primarily for recognition of low-frequency words and another that is low-frequency words and another that is used for the recognition of high- used for the recognition of high- frequency words. frequency words. INTRODUCTION INTRODUCTION Study list: Study list: 80 words were presented, one at a time, 80 words were presented, one at a time, on the computer screen. on the computer screen. Words were either high frequency (mean Words were either high frequency (mean kf = 156.0 ) or low frequency (mean kf = kf = 156.0 ) or low frequency (mean kf = 2.4 ) depending on group assignment. 2.4 ) depending on group assignment. Recognition test: Recognition test: 160 words were presented on the computer 160 words were presented on the computer screen, one at a time. screen, one at a time. Test word frequencies varied across Test word frequencies varied across experiments. experiments. Half of the words appeared in the study Half of the words appeared in the study list (old), and half did not (new). list (old), and half did not (new). Subjects were asked to decide whether Subjects were asked to decide whether the words are old or new, as well as the words are old or new, as well as provide confidence judgments ranging from provide confidence judgments ranging from 3 to 1 (3 = sure, 2 = moderately sure, 1 3 to 1 (3 = sure, 2 = moderately sure, 1 = unsure). = unsure). GENERAL METHOD GENERAL METHOD FREQUENCY EFFECTS IN RECOGNITION MEMORY: FREQUENCY EFFECTS IN RECOGNITION MEMORY: EVIDENCE FOR A MULTIPROCESS DETECTION MODEL EVIDENCE FOR A MULTIPROCESS DETECTION MODEL Marty W. Niewiadomski Marty W. Niewiadomski University of Toronto at Scarborough University of Toronto at Scarborough David G. Smith David G. Smith University of University of Toronto Toronto REFERENCES REFERENCES Banks, W. P. (2000). Recognition and source memory as Banks, W. P. (2000). Recognition and source memory as multivariate decision processes. multivariate decision processes. Psychological Science, 11, Psychological Science, 11, 267-273. 267-273. Glanzer, M. & Adams, J. K. (1985). The mirror effect Glanzer, M. & Adams, J. K. (1985). The mirror effect in recognition memory. in recognition memory. Memory & Cognition, 13, Memory & Cognition, 13, 8-20. 8-20. ACKNOWLEDGEMENTS ACKNOWLEDGEMENTS The authors would like to thank Bill Hockley and Ben The authors would like to thank Bill Hockley and Ben Murdock for the use of their laboratory facilities, as Murdock for the use of their laboratory facilities, as well as their insightful comments and Steve Joordens for well as their insightful comments and Steve Joordens for his helpful suggestions. his helpful suggestions. This research was supported by an NSERC operating grant This research was supported by an NSERC operating grant For additional information please contact David Smith at [email protected] or Marty Niewiadomski at For additional information please contact David Smith at [email protected] or Marty Niewiadomski at [email protected] [email protected] One way to understand these results is One way to understand these results is to assume that the kind of information to assume that the kind of information one uses for recognition decisions is based one uses for recognition decisions is based on the characteristics of the study on the characteristics of the study items. Our results are consistent with the items. Our results are consistent with the idea that when one studies an HF study idea that when one studies an HF study list, recognition decisions are based on list, recognition decisions are based on familiarity (hence the familiarity (hence the frequency effects). frequency effects). studies a list of LF words ones studies a list of LF words ones decisions are based decisions are based on another on another internal variable not based on internal variable not based on familiarity (e.g. familiarity (e.g. Recollection) Recollection) which eliminates frequency which eliminates frequency effects.* effects.* Figures 3a,b,&c demonstrate Figures 3a,b,&c demonstrate how a two- how a two- dimensional model could dimensional model could accommodate the results. When accommodate the results. When one studies one studies only HF words only HF words recognition is based entirely on recognition is based entirely on familiarity. familiarity. When one studies LF When one studies LF words recognition is based on the words recognition is based on the dimension (which we call dimension (which we call Strength Of Recollection). Strength Of Recollection). When one studies a mixed list of When one studies a mixed list of words words recognition is based on a recognition is based on a combination of both dimensions. combination of both dimensions. *Banks (2000) used a similar approach *Banks (2000) used a similar approach to model source to model source memory. memory. DISCUSSION DISCUSSION Legend Legend LL=LF Lure LL=LF Lure LT=LF Target LT=LF Target HL=HF Lure HL=HF Lure HT=HF Target HT=HF Target Figure 3c. Mixed Study Figure 3c. Mixed Study 0 3.5 Hit FA Response Frequency lo hi 0 10 0 10 Yes/No C rit. HL HT LL LT SOF S O R Figure 3a. LF Study Figure 3a. LF Study 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Target LF Lure H F Lure R esponse Frequency 0 10 0 10 SO F SO R Y es/N o C rit. LT HL LL Figure 3b. HF Study Figure 3b. HF Study 0 0.2 0.4 0.6 0.8 1 Target LF Lure H F Lure R esponse Frequency 0 10 0 10 SO F SO R Y es/No C rit. HT HL LL Method: Method: One group studied a list of 80 low One group studied a list of 80 low frequency words and another group studied frequency words and another group studied a list of 80 high-frequency words. At a list of 80 high-frequency words. At test all studied items were presented as test all studied items were presented as probes along with 40 HF lures and 40 LF probes along with 40 HF lures and 40 LF lures. lures. Results: Results: Contrary to the standard within Contrary to the standard within subjects mirror effect subjects mirror effect the Hit rate the Hit rate was not significantly different for was not significantly different for ( ( M M =0.78) and LF ( =0.78) and LF ( M M =0.83) =0.83) targets. For a list of targets. For a list of LF study items LF study items false alarms for HF ( false alarms for HF ( M M =0.13) and =0.13) and LF lures LF lures ( ( M M =0.13) did not differ. =0.13) did not differ. For a list of HF study For a list of HF study items there items there were more FA’s to HF lures ( were more FA’s to HF lures ( M M =0.34) =0.34) than LF lures than LF lures (M (M =0.02) =0.02) . . EXPERIMENT 1 EXPERIMENT 1 Experim ent1 0 0.2 0.4 0.6 0.8 1 LF Study H F Study P(Yes) Target LF Lure H F Lure Method: Method: Same method as Experiment 1 Same method as Experiment 1 except that Lure type except that Lure type (HF or LF) (HF or LF) was blocked. Half the participants was blocked. Half the participants had HF lures had HF lures in the first block and in the first block and the other half had HF the other half had HF lures in the lures in the second. second. Results: Results: Like Experiment 1 Hit rate was Like Experiment 1 Hit rate was the same for HF the same for HF ( ( M=0.77) M=0.77) and LF and LF targets ( targets ( M M =0.75) =0.75) . . For a list of LF For a list of LF study items False study items False alarms for HF alarms for HF ( ( M M =0.24) and LF lures ( =0.24) and LF lures ( M M =0.20) =0.20) did not did not differ. For list of HF study differ. For list of HF study items there were more items there were more False alarms False alarms to HF lures ( to HF lures ( M M =0.39) than LF lures =0.39) than LF lures ( ( M M =0.09). =0.09). Hit rate depended on the type of Hit rate depended on the type of lure for HF but lure for HF but not LF study items. not LF study items. When subjects studied HF items When subjects studied HF items and were tested in and were tested in a context of a context of LF lures the Hit rate was LF lures the Hit rate was higher higher ( ( M M =0.81) then for HF lure context =0.81) then for HF lure context ( ( M M =0.73). =0.73). However, when subjects studied LF However, when subjects studied LF E xperim ent2 H F Study 0 0.2 0.4 0.6 0.8 1 H F Lure LF Lure P(Yes) Hits FA's E xperim ent2 L F Study 0 0.2 0.4 0.6 0.8 H F Lure LF Lure P(Y es) H its FA's EXPERIMENT 2 EXPERIMENT 2

Upload: jason-dalton

Post on 19-Jan-2018

217 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: In a recognition test, participants typically make more hits and fewer false alarms on low-frequency words compared to high frequency words (A pattern

In a recognition test, participants typically make In a recognition test, participants typically make more hits and fewer false alarms on low-frequency words more hits and fewer false alarms on low-frequency words compared to high frequency words (A pattern of results compared to high frequency words (A pattern of results known as the mirror effect, Glanzer & Adams, 1985). In known as the mirror effect, Glanzer & Adams, 1985). In most of these experiments high and low frequency words most of these experiments high and low frequency words are mixed at both study and test. Experiments 1 and 2 are mixed at both study and test. Experiments 1 and 2 show that the hit-rate component of the mirror effect is show that the hit-rate component of the mirror effect is eliminated when participants study a pure list of either eliminated when participants study a pure list of either low or high frequency list. The pattern of false alarms low or high frequency list. The pattern of false alarms also changes depending on the type of list. We suggest also changes depending on the type of list. We suggest that our current pattern of results as well as previous that our current pattern of results as well as previous results can be understood if one assumes that there are results can be understood if one assumes that there are two-dimensions of information available for recognition two-dimensions of information available for recognition decisions: one that is used primarily for recognition of decisions: one that is used primarily for recognition of low-frequency words and another that is used for the low-frequency words and another that is used for the recognition of high-frequency words.recognition of high-frequency words.

INTRODUCTIONINTRODUCTION

Study Study list:list:

•80 words were presented, one at a time, on the computer 80 words were presented, one at a time, on the computer screen.screen.

•Words were either high frequency (mean kf = 156.0 ) or Words were either high frequency (mean kf = 156.0 ) or low frequency (mean kf = 2.4 ) depending on group low frequency (mean kf = 2.4 ) depending on group assignment.assignment.

Recognition test:Recognition test:•160 words were presented on the computer screen, one at 160 words were presented on the computer screen, one at a time.a time.

•Test word frequencies varied across experiments.Test word frequencies varied across experiments.

•Half of the words appeared in the study list (old), and Half of the words appeared in the study list (old), and half did not (new).half did not (new).

•Subjects were asked to decide whether the words are old Subjects were asked to decide whether the words are old or new, as well as provide confidence judgments ranging or new, as well as provide confidence judgments ranging from 3 to 1 (3 = sure, 2 = moderately sure, 1 = unsure).from 3 to 1 (3 = sure, 2 = moderately sure, 1 = unsure).

GENERAL METHODGENERAL METHOD

FREQUENCY EFFECTS IN RECOGNITION MEMORY:FREQUENCY EFFECTS IN RECOGNITION MEMORY:EVIDENCE FOR A MULTIPROCESS DETECTION MODELEVIDENCE FOR A MULTIPROCESS DETECTION MODEL

Marty W. NiewiadomskiMarty W. NiewiadomskiUniversity of Toronto at ScarboroughUniversity of Toronto at Scarborough

David G. SmithDavid G. Smith University of TorontoUniversity of Toronto

REFERENCESREFERENCESBanks, W. P. (2000). Recognition and source memory as multivariate decision Banks, W. P. (2000). Recognition and source memory as multivariate decision

processes. processes. Psychological Science, 11, Psychological Science, 11, 267-273.267-273.

Glanzer, M. & Adams, J. K. (1985). The mirror effect in recognition memory. Glanzer, M. & Adams, J. K. (1985). The mirror effect in recognition memory. Memory & Cognition, 13,Memory & Cognition, 13, 8-20. 8-20.

ACKNOWLEDGEMENTSACKNOWLEDGEMENTSThe authors would like to thank Bill Hockley and Ben Murdock for the use of The authors would like to thank Bill Hockley and Ben Murdock for the use of their laboratory facilities, as well as their insightful comments and Steve their laboratory facilities, as well as their insightful comments and Steve Joordens for his helpful suggestions.Joordens for his helpful suggestions.

This research was supported by an NSERC operating grantThis research was supported by an NSERC operating grant

For additional information please contact David Smith at [email protected] or Marty Niewiadomski at [email protected] additional information please contact David Smith at [email protected] or Marty Niewiadomski at [email protected]

One way to understand these results is to assume that One way to understand these results is to assume that the kind of information one uses for recognition decisions is the kind of information one uses for recognition decisions is based on the characteristics of the study items. Our results based on the characteristics of the study items. Our results are consistent with the idea that when one studies an HF are consistent with the idea that when one studies an HF study list, recognition decisions are based on familiarity study list, recognition decisions are based on familiarity (hence the frequency effects). When one (hence the frequency effects). When one studies a list of LF words ones studies a list of LF words ones decisions are based on another decisions are based on another internal variable not based on internal variable not based on familiarity (e.g. Recollection) which familiarity (e.g. Recollection) which eliminates frequency effects.*eliminates frequency effects.*

Figures 3a,b,&c demonstrate how Figures 3a,b,&c demonstrate how a two-dimensional model could accommodate a two-dimensional model could accommodate the results. When one studies only the results. When one studies only HF words recognition is based entirely HF words recognition is based entirely on familiarity. When one studies LF on familiarity. When one studies LF words recognition is based on the words recognition is based on the second dimension (which we call second dimension (which we call Strength Of Recollection).Strength Of Recollection).When one studies a mixed list of When one studies a mixed list of words recognition is based on a words recognition is based on a combination of both dimensions. combination of both dimensions.

*Banks (2000) used a similar approach to *Banks (2000) used a similar approach to model source memory.model source memory.

DISCUSSIONDISCUSSIONLegendLegend

LL=LF LureLL=LF LureLT=LF TargetLT=LF TargetHL=HF LureHL=HF LureHT=HF TargetHT=HF Target

Figure 3c. Mixed StudyFigure 3c. Mixed Study

0

3.5

Hit FA

Res

pons

e Fr

eque

ncy

lohi

0

10

0 10

Yes/No Crit.

HL

HT

LL

LT

SOF

S O R

Figure 3a. LF StudyFigure 3a. LF Study

00.10.20.30.40.50.60.70.80.9

Target LF Lure HF Lure

Res

pons

e Fr

eque

ncy

0

10

0 10SO F

SOR

Yes/No Crit.

LT

HL

LL

Figure 3b. HF StudyFigure 3b. HF Study

0

0.2

0.4

0.6

0.8

1

Target LF Lure HF Lure

Res

pons

e Fr

eque

ncy

0

10

0 10SO F

SOR

Yes/No Crit.

HT

HL

LL

Method:Method:•One group studied a list of 80 low frequency words and One group studied a list of 80 low frequency words and another group studied a list of 80 high-frequency words. another group studied a list of 80 high-frequency words. At test all studied items were presented as probes along At test all studied items were presented as probes along with 40 HF lures and 40 LF lures. with 40 HF lures and 40 LF lures.

Results:Results:•Contrary to the standard within subjects Contrary to the standard within subjects mirror effect the Hit rate was not mirror effect the Hit rate was not significantly different for HF significantly different for HF ((MM=0.78) and LF (=0.78) and LF (MM=0.83) targets. =0.83) targets. For a list of LF study items false alarms For a list of LF study items false alarms for HF (for HF (MM=0.13) and LF lures =0.13) and LF lures ((MM=0.13) did not differ. For a list of =0.13) did not differ. For a list of HF study items there were more FA’s HF study items there were more FA’s to HF lures (to HF lures (MM=0.34) than LF lures =0.34) than LF lures (M(M=0.02)=0.02)..

EXPERIMENT 1EXPERIMENT 1

Experiment 1

0

0.2

0.4

0.6

0.8

1

LF Study HF Study

P(Y

es)

Target

LF Lure

HF Lure

Method:Method:•Same method as Experiment 1 except Same method as Experiment 1 except that Lure type (HF or LF) was that Lure type (HF or LF) was blocked. Half the participants had HF blocked. Half the participants had HF lures in the first block and the other lures in the first block and the other half had HF lures in the second.half had HF lures in the second.

Results:Results:•Like Experiment 1 Hit rate was the Like Experiment 1 Hit rate was the same for HF (same for HF (M=0.77)M=0.77) and LF targets and LF targets ((MM=0.75)=0.75).. For a list of LF study items For a list of LF study items False alarms for HF (False alarms for HF (MM=0.24) and LF =0.24) and LF lures (lures (MM=0.20) did not differ. For =0.20) did not differ. For list of HF study items there were more list of HF study items there were more False alarms to HF lures (False alarms to HF lures (MM=0.39) than =0.39) than LF lures (LF lures (MM=0.09).=0.09).•Hit rate depended on the type of lure Hit rate depended on the type of lure for HF but not LF study items.for HF but not LF study items.

•When subjects studied HF items and When subjects studied HF items and were tested in a context of LF were tested in a context of LF lures the Hit rate was higher (lures the Hit rate was higher (MM=0.81) =0.81) then for HF lure context (then for HF lure context (MM=0.73).=0.73).•However, when subjects studied LF items, the Hit However, when subjects studied LF items, the Hit rates in the different contexts were the same (rates in the different contexts were the same (MM=0.75)=0.75)

Experiment 2 HF Study

0

0.2

0.4

0.6

0.8

1

HF Lure LF Lure

P(Y

es)

HitsFA's

Experiment 2 LF Study

0

0.2

0.4

0.6

0.8

HF Lure LF Lure

P(Y

es)

Hits

FA's

EXPERIMENT 2EXPERIMENT 2